CN115294135B - Battery edge curling and sealing quality detection method - Google Patents
Battery edge curling and sealing quality detection method Download PDFInfo
- Publication number
- CN115294135B CN115294135B CN202211227365.9A CN202211227365A CN115294135B CN 115294135 B CN115294135 B CN 115294135B CN 202211227365 A CN202211227365 A CN 202211227365A CN 115294135 B CN115294135 B CN 115294135B
- Authority
- CN
- China
- Prior art keywords
- pixel point
- depth
- battery
- region
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/168—Segmentation; Edge detection involving transform domain methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
- G06T7/45—Analysis of texture based on statistical description of texture using co-occurrence matrix computation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/10—Energy storage using batteries
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Geometry (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of image processing, in particular to a battery curling sealing quality detection method, which comprises the steps of obtaining an RGB image, a depth image, a first side view and a second side view of a battery to be detected; acquiring a sealing edge in an RGB image to calculate a complete index, carrying out Hough circle detection on a depth image to obtain an inner circle and an outer circle, and obtaining a region to be generated and a region with known depth based on the inner circle and the outer circle; obtaining a smooth index of each pixel point in the depth known region according to the complete index; respectively obtaining ROI areas in the first side view and the second side view, respectively obtaining a transverse gradient value and a longitudinal gradient value of each pixel point based on the ROI areas to obtain a depth gradient fusion value, and further obtaining a depth gradient image of a region to be generated; and acquiring a cap depth image by using the depth gradient image, and combining the RGB image, the depth image and the cap depth image to obtain the quality defect type of the battery to be detected. The invention improves the quality detection junction of the battery seal.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a battery edge curling and sealing quality detection method.
Background
The cylindrical lithium battery is an important battery type in the field of electric vehicles, thousands of cylindrical lithium batteries can form a power supply module of the electric vehicle, common cylindrical lithium batteries such as 4680 batteries of Tesla, and the packaging process is crucial in the production process. Because the circumstances such as wearing and tearing of punching press machine, electric core production process's error probably takes place the dislocation, presses excessively to cause the turn-up to seal the defect, then probably take place trouble such as swell short circuit in subsequent use, can lead to the casualties accident seriously.
The method for detecting the curling sealing defect of the battery at present is to shoot a lithium battery image without the curling sealing defect by utilizing x-rays, take the lithium battery image as an image template, and then carry out image matching on the image of the lithium battery to be detected and the image template so as to detect the curling sealing defect of the lithium battery to be detected.
Disclosure of Invention
In order to solve the above technical problems, an object of the present invention is to provide a method for detecting a quality of a battery curled edge seal, which adopts the following technical scheme:
the method comprises the steps of obtaining an RGB image, a depth image, a first side view and a second side view of a battery to be detected, and rotating the battery to be detected by 90 degrees after the second side view is collected from the first side view;
obtaining a sealing edge in the RGB image, and comparing the standard sealing edge in the RGB image corresponding to the battery without the quality defect to obtain a complete index of the sealing edge; carrying out Hough circle detection on the depth image to obtain an inner circle and an outer circle, taking a region between the inner circle and the outer circle as a region to be generated, and taking a region corresponding to the inner circle as a depth known region; acquiring a circular neighborhood of each pixel point in a depth known region according to the complete index, and respectively calculating a gray level co-occurrence matrix in the circular neighborhood of each pixel point to obtain a corresponding smooth index;
respectively acquiring ROI (region of interest) areas in a first side view and a second side view, respectively acquiring corresponding target positions of corresponding pixel points in the ROI areas according to coordinates of the pixel points in the areas to be generated, adaptively acquiring a rectangular window by taking the target positions as centers, and respectively acquiring transverse gradient values and longitudinal gradient values of the corresponding pixel points according to gray values of pixel points in each row in the rectangular window; fusing the transverse gradient value and the longitudinal gradient value of each pixel point in the region to be generated to obtain a depth gradient fusion value of the corresponding pixel point, and replacing the original gray value of the pixel point with the depth gradient fusion value to obtain a depth gradient map of the region to be generated;
inputting the depth image and the depth gradient map into a generation countermeasure network to obtain a cap depth image; and inputting the RGB image, the depth image and the cap depth image of the battery to be detected into a quality detection network to obtain the quality defect type of the battery to be detected.
Further, the method for obtaining the integrity index of the edge of the seal includes:
and taking the number of the edge pixel points on the sealing edge as a numerator and the number of the edge pixel points on the standard sealing edge as a denominator to obtain a corresponding ratio, and taking the ratio as a complete index of the sealing edge.
Further, the method for acquiring the circular neighborhood of each pixel point comprises the following steps:
and inputting the ratio between the preset proportionality coefficient and the complete index into a natural logarithm to obtain the radius of the circular neighborhood, and obtaining the circular neighborhood of the corresponding pixel point according to the radius by taking the pixel point as the center of a circle.
Further, the method for obtaining the smoothing index includes:
and acquiring entropy and contrast of the gray level co-occurrence matrix, calculating a product between a second proportional coefficient and the contrast, and taking the reciprocal of the sum of the product and the entropy as a smooth index.
Further, the method for obtaining the transverse gradient value includes:
acquiring a connecting line between a current pixel point and the center of an inner circle in a region to be generated, and calling a pixel point belonging to a region with known depth on the connecting line as a correlation point of the current pixel point;
determining a target position of a current pixel point in a first side view according to the abscissa of the current pixel point, and acquiring a corresponding rectangular window in an ROI (region of interest) by taking the target position as a center, wherein the rectangular window is obtained according to the smooth index of the associated point of the current pixel point and the height of the ROI;
respectively calculating the gray value mean value of each row of pixel points in the rectangular window, and taking the difference absolute value of the gray value mean value of the ith row of pixel points and the gray value mean value of the (i + 1) th row of pixel points as the local gray value change value of the ith row of pixel points; and accumulating the local gray scale change values of other rows except the last row in the rectangular window to obtain the transverse gradient value of the current pixel point.
Further, the method for acquiring the longitudinal gradient value includes:
acquiring a connecting line between a current pixel point and the center of an inner circle in a region to be generated, and calling a pixel point belonging to a region with known depth on the connecting line as a correlation point of the current pixel point;
determining a target position of the current pixel point in the second side view according to the ordinate of the current pixel point, and acquiring a corresponding rectangular window in the ROI area by taking the target position as a center, wherein the rectangular window is obtained according to the smooth index of the associated point of the current pixel point and the height of the ROI area;
respectively calculating the gray value mean value of each row of pixel points in the rectangular window, and taking the difference absolute value of the gray value mean value of the ith row of pixel points and the gray value mean value of the (i + 1) th row of pixel points as the local gray value change value of the ith row of pixel points; and accumulating the local gray scale change values of other rows except the last row in the rectangular window to obtain the longitudinal gradient value of the current pixel point.
Further, the method for obtaining the depth gradient fusion value includes:
the longitudinal gradient value and the transverse gradient value are respectively squared, and then the results of the addition and the re-evolution are taken as depth gradient fusion values.
The embodiment of the invention at least has the following beneficial effects: the method comprises the steps of representing the characteristics of a battery to be detected from different angles by acquiring an RGB image, a depth image, a first side view and a second side view of the battery to be detected, and improving the identification degree of a sealing position of the battery; obtaining a sealing edge in the RGB image, and calculating a complete index of the sealing edge by comparing the standard sealing edge of the battery without quality defects; carrying out Hough circle detection on the depth image to obtain an inner circle and an outer circle, obtaining a region to be generated corresponding to the edge of the seal and a depth known region corresponding to the inner circle from the inner circle and the outer circle, and obtaining a smooth index of each pixel point in the depth known region corresponding to the circular neighborhood based on the complete index, wherein the larger the smooth index is, the smaller the non-fitting degree of the cap of the battery is; since the lateral view can reflect the abscissa characteristic and the ordinate characteristic of each coordinate point in the top view, the transverse gradient value and the longitudinal gradient value of each pixel point in the region to be generated are respectively calculated based on the ROI in the first lateral view and the second lateral view, the depth gradient fusion value of the corresponding pixel point is obtained through fusion, and then the depth gradient map of the region to be generated can be obtained, so that the depth information of the invisible region of the battery is accurately restored; the depth image and the depth gradient image are combined, and a complete battery cap image is obtained by utilizing the generated countermeasure network, so that the subsequent quality detection is facilitated; the RGB image, the depth image and the cap depth image of the battery to be detected are input into a quality detection network, and an accurate quality detection result of the battery seal can be obtained.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart illustrating steps of a method for detecting a quality of a battery crimp seal according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a result of hough circle detection performed on a depth image in the embodiment of the present invention;
FIG. 3 is a diagram illustrating a ROI area in an embodiment of the present invention;
fig. 4 is an operation schematic diagram of a pixel point O of a region to be generated in the embodiment of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description is provided for the battery crimp sealing quality detection method according to the present invention, and the specific implementation manner, structure, features and effects thereof are described in the following with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The specific scenes aimed by the method are as follows: a 18650 sealing machine is selected to seal the cylindrical lithium battery, and for example, the detection of a certain cylindrical lithium battery is taken as an example, in the sealing process, an arc-shaped surface is arranged at the curling position, and the first arc-shaped surfaces of the plurality of clamps form an annular holding surface for holding and releasing the cylindrical lithium battery; the bottom of the crimping upper die is provided with a crimping inner rotary wall, the inner diameter of the crimping inner rotary wall is gradually reduced from bottom to top, and the crimping upper die acts on the cylindrical lithium battery and realizes crimping during the up-down movement process of the crimping inner rotary wall. Various sealing defects are easy to occur in the sealing process of the cylindrical lithium battery, such as: (1) Surface breakage, the smooth overhang of the bead depending lip under the normal bead, a defect that usually occurs at the edge seal of three-piece cans because the thickness is greater here than elsewhere; (2) Sharp edges appear, the operation pressure of the first and/or second seaming rollers is too tight, the residue of the contents exists at the seaming position, the pressure of the tin supporting rotary table is too high, and the seaming rollers and/or the chuck are worn, so that the sharp edges appear at the upper part of the inner side of the seal; (3) The phenomenon that the curled edge is broken, generally, the curled edge is too tight due to the over-tight pressure of the first seaming roller and the second seaming roller, so that the curled edge part of the outer layer is broken, and the solution is that the pressure of the two seaming rollers is properly released; (4) Untight seals (false roll) are a serious defect in the rolling process, which can lead to leakage of the liquid in the cell; (5) And incomplete crimping, incomplete crimping or crimping defect caused by over-loose pressure at part of the place in the crimping process, so that the quality of the crimped seal of the cylindrical lithium battery is detected.
The following describes a specific scheme of the battery crimp sealing quality detection method provided by the invention in detail with reference to the accompanying drawings.
Referring to fig. 1, a flow chart illustrating steps of a method for detecting a quality of a battery crimp seal according to an embodiment of the present invention is shown, the method including the following steps:
and S001, acquiring an RGB image, a depth image, a first side view and a second side view of the battery to be detected, wherein the second side view is obtained by rotating the battery to be detected by 90 degrees after the first side view is acquired.
Specifically, this scheme utilization RGB camera, laser depth camera, industry X-ray machine carry out the image acquisition of battery, promptly when the battery conveys the detection position of industry X-ray machine, RGB camera and laser depth camera are from overlooking visual angle and shoot the battery, obtain RGB image A and depth image B, however industry X-ray machine is from looking sideways at visual angle and shoot the battery, obtain first side view C, after having shot first side view, with the battery according to vertical axis clockwise rotation 90 degrees, carry out the shooting of second time and obtain second side view D.
It should be noted that, the battery is subjected to perspective detection by using an industrial X-ray machine in the production process, which is a common quality detection means, and the defects therein are found out through a certain magnification, so that 100% detection can be achieved, and not only can the positions of the positive electrode and the negative electrode be detected, and whether the stacked roll cores are neat or not be detected, but also whether the curled edge sealing has defects or not can be detected.
The purpose of taking the second side view D by rotating the battery by 90 degrees is: although industry X-ray machine from overlooking visual angle shooting can be complete see the shape that the round was sealed the department, but there is the interference of the inside book core of battery, X-ray's penetrability is very poor, lead to the discernment degree poor to sealing the department, consequently, it can more directly perceived the embodiment battery internal conditions to wait to detect the side view of battery from looking sideways at the visual angle shooting, strengthen the discernment degree of sealing the department, and through obtaining the side view that the shooting angle differs 90 degrees, can be used for the degree of depth image that the angle was shot is overlooked to the sign and use every pixel point to construct the rectangular coordinate system as the initial point, utilize first side view to reflect the characteristic of pixel on the X axle, second side view reflects the characteristic of pixel on the y axle.
S002, obtaining a sealing edge in the RGB image, and comparing the standard sealing edge in the RGB image corresponding to the battery without the quality defect to obtain a complete index of the sealing edge; carrying out Hough circle detection on the depth image to obtain an inner circle and an outer circle, taking a region between the inner circle and the outer circle as a region to be generated, and taking a region corresponding to the inner circle as a depth known region; and acquiring a circular neighborhood of each pixel point in the known depth region according to the complete index, and respectively calculating a gray level co-occurrence matrix in the circular neighborhood of each pixel point to obtain a corresponding smooth index.
Specifically, based on the method of step S001, an RGB image A of a battery free from quality defects is obtained ′ Detecting the sealing edge of the battery by using Hough circle detection to respectively obtain RGB images A ′ And calculating the number of edge pixel points on each seal edge optimal circle corresponding to the seal edge in the RGB image A, and taking the RGB image A as a reference ′ The number of the edge pixel points on the corresponding sealing edge optimal circle is denominator, the number of the edge pixel points on the sealing edge optimal circle corresponding to the RGB image A is numerator to obtain a corresponding ratio, and then the ratio is used as a complete index Full of the sealing edge of the battery to be detected.
It should be noted that the hough circle detection method is a known technique, and the scheme is not described again.
Using the hough circle detection method for the depth image B, the outer edge line 1 and the inner edge line 2 shown in fig. 2 can be obtained, where the outer edge line 1 corresponds to the outer circle and the inner edge line 2 corresponds to the inner circle. And taking the region between the outer edge line and the inner edge line as a region to be generated, namely a shielding region, wherein the region corresponding to the inner edge line is a region with known depth.
If the area to be generated changes, the area with known depth can display some features, so that the depth gradient map E of the area to be generated is obtained according to the depth value of each pixel point in the area with known depth.
Acquiring the radius of a circular neighborhood of each pixel point in a region with known depth according to the complete index Full of the sealing edge of the battery to be detected, wherein the radius of the circular neighborhood
Wherein ln is natural logarithm, k 1 Is a coefficient of proportionality that is,empirical value k 1 =0.05. The smaller the value of the complete index Full is, the less clear the inner edge line is, the more likely an unsmooth area exists in the corresponding area with the known depth, and the larger the perception domain of the pixel point needs to be enlarged, the larger the corresponding R value is.
Acquiring a circular neighborhood of each pixel point in the known depth region based on the radius R, acquiring a gray level co-occurrence matrix corresponding to the circular neighborhood according to the depth value of each pixel point, calculating the entropy ENT and the contrast CON of the gray level co-occurrence matrix, and calculating the smoothness index Smo of the sensing domain of the corresponding pixel point in the known depth region by combining the entropy ENT and the contrast CON of the gray level co-occurrence matrix:
wherein k is 2 The empirical value is k, which is a proportionality coefficient, according to the smooth response capability of the entropy ENT and the contrast CON 2 ＝0.8。
The entropy ENT reflects the intensity and the non-uniform degree of depth change, the larger the entropy ENT is, the more probable the non-fitting condition of the cap of the battery to be detected exists, the smaller the smoothness index in the circular neighborhood of the corresponding pixel point is, the greater the contrast CON reflects the maximum degree of depth change, the larger the contrast CON is, the more probable the non-fitting degree of the cap of the battery to be detected is, and the smaller the smoothness index in the circular neighborhood of the corresponding pixel point is.
And acquiring the smooth index in the circular neighborhood of each pixel point in the region with known depth by using the smooth index acquisition method.
Step S003, obtaining ROI areas in the first side view and the second side view respectively, obtaining corresponding target positions of corresponding pixel points in the ROI areas respectively according to coordinates of the pixel points in the areas to be generated, obtaining a rectangular window in a self-adaptive mode by taking the target positions as centers, and obtaining transverse gradient values and longitudinal gradient values of the corresponding pixel points respectively according to gray values of pixel points in each row in the rectangular window; and fusing the transverse gradient value and the longitudinal gradient value of each pixel point in the region to be generated to obtain a depth gradient fusion value of the corresponding pixel point, and replacing the original gray value of the pixel point with the depth gradient fusion value to obtain a depth gradient image of the region to be generated.
Specifically, as shown in fig. 3, an ROI region, i.e., a rectangular region corresponding to a dashed frame in the drawing, is divided in the first side view C and the second side view D, where the ROI region is a minimum circumscribed rectangle corresponding to a portion above an upper boundary of a sidewall groove of the cell.
And obtaining a depth gradient map E of the region to be generated by combining the gray value of each pixel point in the ROI region in the first side view C, the gray value of each pixel point in the ROI region in the second side view D and the smoothness index in the circular neighborhood of each pixel point in the depth known region in the depth image B, wherein the details are as follows:
as shown in fig. 4, any one pixel point O in the region to be produced is selected, the center of the inner circle in the depth image B and the pixel point O are connected to obtain a dotted line, a rectangular coordinate system is constructed with the pixel point O as an origin, two vertical arrows in the drawing represent the rectangular coordinate system, and the x axis and the longitudinal y axis are represented in a transverse manner.
Since the first side view C can reflect the characteristics of the pixel point O in the to-be-generated region on the x axis, and the second side view D can reflect the characteristics of the pixel point O in the to-be-generated region on the y axis, the associated point of the pixel point O is selected in the to-be-generated region, that is, the pixel point on the dotted line corresponding to the pixel point O and belonging to the region with known depth is referred to as the associated point.
Determining the position O of the pixel point O in the first side view C according to the abscissa of the pixel point O ′ In the position O ′ A rectangular window is developed for the center in the ROI area of the first side view C, the size of the rectangular window depends on the smoothing index Smo of the associated point, the height H of the rectangular window is equal to the height of the ROI area, and the width of the matrix window
Wherein N is the number of association points, smo i Is the smoothing index of the ith associated point.
It should be noted that the smoother the neighborhood region, the lower the probability of defect existence, the smaller the rectangular window size, and the more careful the extracted features, so the larger the sum of the smoothness indexes of the associated points, the smoother the neighborhood region, and the smaller the width of the corresponding rectangular window.
Analyzing each row of pixel points in the matrix window of the pixel point O, and respectively calculating the local gray scale change value of each row of pixel points: taking the ith row of pixels as an example, the gray value mean of each row of pixels is respectively calculated, the absolute value of the difference between the gray value mean of the ith row of pixels and the gray value mean of the (i + 1) th row of pixels is used as the local gray value change value of the ith row of pixels, and the local gray value change value reflects the bending degree of the cap of the area to be generated in the first side view C.
Since the last row in the rectangular window cannot analyze the local gray level variation value, the local gray level variation values of the rows except the last row are accumulated to obtain the transverse gradient value D of the pixel point O x 。
Similarly, the position O 'of the pixel point O in the second side view D is determined according to the ordinate of the pixel point O, a rectangular window is expanded in the ROI area of the second side view D by taking the position O' as the center, the size of the rectangular window depends on the smoothness index Smo of the associated point, the height H of the rectangular window is equal to the height of the ROI area, and the width of the matrix window
Wherein N is the number of association points, smo i A smoothing index of the ith associated point; calculating the local gray variation value of each row of pixel points in the rectangular window to obtain the longitudinal gradient value D of the pixel point O y 。
Transverse gradient value D to pixel point O x And longitudinal gradient values D y Gradient value fusion is carried out to obtain a depth gradient fusion value of the prime point O
Further fusing the obtained depth gradient with a value D merg And replacing the pixel point O to obtain the original gray value.
Fusing values D according to depth gradient merg The method for obtaining the depth gradient map E comprises the steps of obtaining the depth gradient fusion value of each pixel point in the region to be generated, and replacing the original gray value of each pixel point with the depth gradient fusion value, so as to obtain the depth gradient map E of the region to be generated.
Step S004, inputting the depth image and the depth gradient map into a generation countermeasure network to obtain a cap depth image; and inputting the RGB image, the depth image and the cap depth image of the battery to be detected into a quality detection network to obtain the quality defect type of the battery to be detected.
Specifically, a Cylinder-Battery-GAN generation countermeasure network is constructed and trained: cylinder-Battery represents a cylindrical lithium Battery, a GAN generation countermeasure network is a neural network type, the specific structure of the Cylinder-Battery-GAN generation countermeasure network is a structure of a DC-GAN generation countermeasure network, but the input of the Cylinder-Battery-GAN generation countermeasure network is different from the input of the DC-GAN generation countermeasure network in the scheme.
Wherein, the training process of the Cylinder-Battery-GAN to generate the confrontation network is as follows: collecting RGB images, depth images, first side views and second side views of a plurality of batteries by using the step S001, and then arranging a big data annotation expert to perform defect type annotation on the images, wherein the defect types comprise normal 01, untight 02 sealing and surface damage 03; acquiring a unsealing depth image G corresponding to each battery, namely after the batteries of each defect type corresponding to the battery of the normal 01, the untight 02 seal and the surface damage 03 are subjected to image acquisition and image marking, disassembling the edge-curling sealing position of the battery, shooting an unobstructed depth image again by using a depth camera to be used as the unsealing depth image G, and further obtaining a target generation image set, wherein 80% of the image set is used as a training set, and 20% of the image set is used as a test set for training; and (3) performing reverse error propagation by using a cross entropy loss function, and finally obtaining a Cylinder-Battery-GAN generation countermeasure network with a good generation effect by using Adam by using the optimizer.
It should be noted that the unpacking depth image G of a normal battery is just that one normal battery needs to be disassembled, and is further used as the unpacking depth image G of all normal batteries.
The generation principle of the Cylinder-Battery-GAN generation countermeasure network is the countermeasure of a generator and a discriminator, and is specifically represented as follows: the function of the generator is to generate a cap depth image F according to the input depth image B, the depth gradient map E and data, and the input of the discriminator is the cap depth image F generated by the generator and the acquired unsealing depth image G; the function of the discriminator is to determine whether the generated image is intended by the user, i.e. true or false, and to back-propagate the error, optimizing the parameters of the generator. The generation process of the generator is unordered, the cap depth image F generated by the generator more and more meets the required unsealing depth image G along with the training until the probability that the discriminator cannot judge whether the image is true or false, or the probability that the image is true or false is 0.5, nash balance is achieved, and the game is finished at the moment.
Utilizing trained Cylinder-Battery-GAN to generate a countermeasure network, and acquiring a cap depth image F of the Battery to be detected: in the using process, the restoration of the region to be generated can be realized only by using the trained generator, namely the input of the generator is a depth image B and a depth gradient image E, the feature extraction is completed, and the output is a cap depth image F.
Constructing a quality detection network, wherein the network structure is ResNet50, inputting the RGB image A, the depth image B and the cap depth image F of the battery to be detected into the quality detection network, and outputting the quality defect type of the battery seal, namely outputting the quality defect type comprising a plurality of target frames and corresponding defect types in the frames: the seal is not tight 02 or the surface is broken 03.
Note that the RGB image a can reflect a region with a damaged surface 03, and the depth image B and the cap depth image F can reflect a region with a not-tight seal 02 according to a depth difference between a region to be generated in the cap depth image F and the depth image B.
Further, according to the quality defect type detected by the battery seal, different treatments can be subsequently carried out on batteries with different defect types: the battery with the untight seal 02 needs to be reworked and sealed again, and the battery with the damaged surface 03 can directly repair and reinforce the appearance.
In summary, in the embodiment of the present invention, the RGB image, the depth image, the first side view and the second side view of the battery to be detected are obtained; acquiring a sealing edge in an RGB image and a standard sealing edge of a battery without quality defects, and comparing and calculating a complete index of the sealing edge; carrying out Hough circle detection on the depth image to obtain an inner circle and an outer circle, acquiring a region to be generated corresponding to a sealing edge and a depth known region corresponding to the inner circle from the inner circle and the outer circle, and acquiring a smooth index of each pixel point in the depth known region corresponding to a circular neighborhood based on a complete index; respectively calculating the transverse gradient value and the longitudinal gradient value of each pixel point in the region to be generated based on the ROI (region of interest) in the first side view and the second side view, obtaining the depth gradient fusion value of the corresponding pixel point through fusion, and further obtaining the depth gradient map of the region to be generated, so that the depth information of the invisible region of the battery is accurately restored; the depth image and the depth gradient image are combined, and a complete battery cap image is obtained by utilizing the generated countermeasure network, so that the subsequent quality detection is facilitated; the RGB image, the depth image and the cap depth image of the battery to be detected are input into a quality detection network, and an accurate quality detection result of the battery seal can be obtained.
It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. And specific embodiments thereof have been described above. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
All the embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that are within the spirit of the present invention are intended to be included therein.
Claims (5)
1. A battery edge curling and sealing quality detection method is characterized by comprising the following steps:
the method comprises the steps of obtaining an RGB image, a depth image, a first side view and a second side view of a battery to be detected, and rotating the battery to be detected by 90 degrees after the second side view is collected from the first side view;
obtaining a sealing edge in the RGB image, and comparing the standard sealing edge in the RGB image corresponding to the battery without the quality defect to obtain a complete index of the sealing edge; carrying out Hough circle detection on the depth image to obtain an inner circle and an outer circle, taking a region between the inner circle and the outer circle as a region to be generated, and taking a region corresponding to the inner circle as a depth known region; acquiring a circular neighborhood of each pixel point in a depth known region according to the complete index, and respectively calculating a gray level co-occurrence matrix in the circular neighborhood of each pixel point to obtain a corresponding smooth index;
respectively acquiring ROI (region of interest) areas in a first side view and a second side view, respectively acquiring corresponding target positions of corresponding pixel points in the ROI areas according to coordinates of the pixel points in the areas to be generated, adaptively acquiring a rectangular window by taking the target positions as centers, and respectively acquiring transverse gradient values and longitudinal gradient values of the corresponding pixel points according to gray values of pixel points in each row in the rectangular window; fusing the transverse gradient value and the longitudinal gradient value of each pixel point in the region to be generated to obtain a depth gradient fusion value of the corresponding pixel point, and replacing the original gray value of the pixel point with the depth gradient fusion value to obtain a depth gradient map of the region to be generated;
inputting the depth image and the depth gradient map into a generation countermeasure network to obtain a cap depth image; inputting the RGB image, the depth image and the cap depth image of the battery to be detected into a quality detection network to obtain the quality defect type of the battery to be detected;
the method for acquiring the complete index of the edge of the seal comprises the following steps:
taking the number of edge pixel points on the edge of the seal as a numerator and the number of edge pixel points on the edge of the standard seal as a denominator to obtain a corresponding ratio, and taking the ratio as a complete index of the seal edge;
the method for acquiring the smoothing index comprises the following steps:
acquiring entropy and contrast of the gray level co-occurrence matrix, calculating a product between a second proportional coefficient and the contrast, and taking the reciprocal of the sum of the product and the entropy as a smooth index;
the ROI area refers to a minimum circumscribed rectangle corresponding to a part above the upper boundary of a groove on the side wall of the battery;
the method for respectively acquiring the corresponding target positions of the corresponding pixel points in the ROI according to the coordinates of each pixel point in the region to be generated comprises the following steps:
determining the target position of the corresponding pixel point in the ROI area of the first side view according to the abscissa of each pixel point in the area to be generated; and determining the target position of the corresponding pixel point in the ROI area of the second side view according to the vertical coordinate of each pixel point in the area to be generated.
2. The method for detecting the quality of the battery curled edge sealing according to claim 1, wherein the method for obtaining the circular neighborhood of each pixel point comprises the following steps:
and inputting the ratio between the preset proportionality coefficient and the complete index into a natural logarithm to obtain the radius of the circular neighborhood, and obtaining the circular neighborhood of the corresponding pixel point according to the radius by taking the pixel point as the center of a circle.
3. The method for detecting the quality of the battery crimp seal according to claim 1, wherein the method for acquiring the transverse gradient value comprises the following steps:
acquiring a connecting line between a current pixel point and the center of an inner circle in a region to be generated, and calling a pixel point belonging to a region with known depth on the connecting line as a correlation point of the current pixel point;
determining a target position of a current pixel point in a first side view according to the abscissa of the current pixel point, and acquiring a corresponding rectangular window in an ROI (region of interest) by taking the target position as a center, wherein the rectangular window is obtained according to the smooth index of the associated point of the current pixel point and the height of the ROI;
respectively calculating the gray value mean value of each row of pixel points in the rectangular window, and taking the difference absolute value of the gray value mean value of the ith row of pixel points and the gray value mean value of the (i + 1) th row of pixel points as the local gray value change value of the ith row of pixel points; and accumulating the local gray scale change values of other rows except the last row in the rectangular window to obtain the transverse gradient value of the current pixel point.
4. The method for detecting the quality of the battery crimp seal according to claim 1, wherein the method for acquiring the longitudinal gradient value comprises the following steps:
acquiring a connecting line between a current pixel point and the center of an inner circle in a region to be generated, and calling a pixel point belonging to a region with known depth on the connecting line as a correlation point of the current pixel point;
determining a target position of the current pixel point in the second side view according to the ordinate of the current pixel point, and acquiring a corresponding rectangular window in the ROI area by taking the target position as a center, wherein the rectangular window is obtained according to the smooth index of the associated point of the current pixel point and the height of the ROI area;
respectively calculating the gray value mean value of each row of pixel points in the rectangular window, and taking the difference absolute value of the gray value mean value of the ith row of pixel points and the gray value mean value of the (i + 1) th row of pixel points as the local gray value change value of the ith row of pixel points; and accumulating the local gray scale change values of other lines except the last line in the rectangular window to obtain the longitudinal gradient value of the current pixel point.
5. The method for detecting the quality of the battery crimp seal according to claim 1, wherein the method for obtaining the depth gradient fusion value comprises the following steps:
the longitudinal gradient value and the transverse gradient value are respectively squared, and then the results of the addition and the re-evolution are taken as depth gradient fusion values.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211227365.9A CN115294135B (en) | 2022-10-09 | 2022-10-09 | Battery edge curling and sealing quality detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211227365.9A CN115294135B (en) | 2022-10-09 | 2022-10-09 | Battery edge curling and sealing quality detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115294135A CN115294135A (en) | 2022-11-04 |
| CN115294135B true CN115294135B (en) | 2023-01-03 |
Family
ID=83833936
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211227365.9A Active CN115294135B (en) | 2022-10-09 | 2022-10-09 | Battery edge curling and sealing quality detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115294135B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116542967B (en) * | 2023-06-29 | 2023-10-03 | 厦门微图软件科技有限公司 | Method, device and equipment for detecting defects of lithium battery pole |
| CN117152088B (en) * | 2023-09-01 | 2024-02-06 | 北京奥乘智能技术有限公司 | Method, device, equipment and storage medium for detecting seal of medicine package |
| CN117726574B (en) * | 2023-09-13 | 2024-04-26 | 东莞市言科新能源有限公司 | Packaging system and method for producing polymer lithium ion battery |
| CN117862891B (en) * | 2024-03-11 | 2024-06-28 | 深圳市美迪泰克医药有限公司 | Production process of electronic booster |
| CN118761014A (en) * | 2024-09-05 | 2024-10-11 | 浙江利强包装科技有限公司 | Vacuum machine packing tightness detection system based on convolutional neural network |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105548216A (en) * | 2016-01-15 | 2016-05-04 | 浙江野马电池有限公司 | Visual detection method for appearance of semi-finished battery |
| CN113313677A (en) * | 2021-05-17 | 2021-08-27 | 武汉工程大学 | Quality detection method for X-ray image of wound lithium battery |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105552259B (en) * | 2016-02-22 | 2019-03-12 | 浙江野马电池有限公司 | A kind of groove edge curling seal device and edge curling seal method |
| CN109598721B (en) * | 2018-12-10 | 2021-08-31 | 广州市易鸿智能装备有限公司 | Defect detection method and device for battery pole piece, detection equipment and storage medium |
| CN110349120A (en) * | 2019-05-31 | 2019-10-18 | 湖北工业大学 | Solar battery sheet detection method of surface flaw |
| CN111105405B (en) * | 2019-12-24 | 2020-12-25 | 芜湖楚睿智能科技有限公司 | New energy lithium battery surface defect detection method based on adaptive deep learning |
| JP7393313B2 (en) * | 2020-09-29 | 2023-12-06 | タカノ株式会社 | Defect classification device, defect classification method and program |
| CN114972258B (en) * | 2022-05-27 | 2023-04-07 | 深圳先进技术研究院 | Battery surface defect detection method and system based on machine vision and related equipment |
| CN115147370A (en) * | 2022-06-30 | 2022-10-04 | 章鱼博士智能技术(上海)有限公司 | Battery top cover welding defect detection method and device, medium and electronic equipment |
-
2022
- 2022-10-09 CN CN202211227365.9A patent/CN115294135B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105548216A (en) * | 2016-01-15 | 2016-05-04 | 浙江野马电池有限公司 | Visual detection method for appearance of semi-finished battery |
| CN113313677A (en) * | 2021-05-17 | 2021-08-27 | 武汉工程大学 | Quality detection method for X-ray image of wound lithium battery |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115294135A (en) | 2022-11-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115294135B (en) | Battery edge curling and sealing quality detection method | |
| CN111307818B (en) | Visual online detection method for laser welding spot of lithium battery tab | |
| CN111814867B (en) | Training method of defect detection model, defect detection method and related device | |
| CN111223088B (en) | Casting surface defect identification method based on deep convolutional neural network | |
| CN113658132B (en) | Computer vision-based structural part weld joint detection method | |
| CN112241699A (en) | Object defect category identification method and device, computer equipment and storage medium | |
| CN114594114B (en) | Full-automatic online nondestructive testing method for lithium battery cell | |
| CN109740515B (en) | Evaluation method and device | |
| CN104537650B (en) | Polarity direction determination method and device for aluminum electrolytic capacitor | |
| CN114581446B (en) | Battery core abnormity detection method and system of laminated battery | |
| CN115314714A (en) | Data compression method for weld image storage | |
| CN116523921B (en) | Detection method, device and system for tab turnover condition | |
| CN115018846B (en) | AI intelligent camera-based multi-target crack defect detection method and device | |
| CN116503348B (en) | Method and equipment for detecting alignment degree of cathode and anode plates of battery core of coiled lithium battery | |
| CN115147370A (en) | Battery top cover welding defect detection method and device, medium and electronic equipment | |
| CN112132196A (en) | Cigarette case defect identification method combining deep learning and image processing | |
| CN115205198A (en) | Method, system and equipment for detecting quality of tab and storage medium | |
| CN116051542B (en) | Defect detection method and defect detection device | |
| CN118096687A (en) | Yolov 8-based substation equipment defect detection method | |
| CN117314880A (en) | Image defect detection method and device | |
| CN115147386A (en) | Defect detection method and device for U-shaped pipe and electronic equipment | |
| CN116342589A (en) | Cross-field scratch defect continuity detection method and system | |
| CN116128833A (en) | Weld defect detection method and device, readable storage medium and electronic equipment | |
| CN112364783B (en) | Part detection method and device and computer readable storage medium | |
| CN114092441A (en) | Product surface defect detection method and system based on dual neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230310 Address after: 276000 intersection of Changshen line and Gongye 1st Road, Tangtou street, Hedong District, Linyi City, Shandong Province Patentee after: Shandong Huatai New Energy Battery Co.,Ltd. Address before: 276032 Intersection of Shusheng Street and Huatai Street, Tangtou Town, Hedong District, Linyi City, Shandong Province Patentee before: LINYI HUATAI BATTERY Co.,Ltd. |
|
| TR01 | Transfer of patent right |