CN112686152A - Crop pest and disease identification method with multi-size input and multi-size targets - Google Patents
Crop pest and disease identification method with multi-size input and multi-size targets Download PDFInfo
- Publication number
- CN112686152A CN112686152A CN202011606979.9A CN202011606979A CN112686152A CN 112686152 A CN112686152 A CN 112686152A CN 202011606979 A CN202011606979 A CN 202011606979A CN 112686152 A CN112686152 A CN 112686152A
- Authority
- CN
- China
- Prior art keywords
- ratio
- test
- train
- sliding window
- identification
- 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.)
- Granted
Links
Images
Classifications
-
- 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
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
- Y02A40/10—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in agriculture
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a crop pest and disease identification method with multi-size input and multi-size targets, which relates to the technical field of image identification, and is characterized in that original crop pest and disease images are collected and target marking is carried out; and converting the marked crop disease and pest image into a data set, wherein the data set is divided into: a training set, a verification set and a test set; constructing a target detection network; training and verifying the target detection network through a training set and a verification set; the method comprises the steps of carrying out model reasoning on a target detection network after test set input training and verification to obtain a first reasoning result, carrying out sliding window recognition, filling recognition or non-maximum suppression processing on the first reasoning result, further detecting leaves and fruits of plant diseases and insect pests in the first reasoning result to obtain a final recognition result, optimizing a model reasoning link, and effectively improving the recall rate and the accuracy rate of the model to a multi-size input and multi-size target detection model under the condition of not increasing the labeling cost so as to improve the crop plant disease and insect pest recognition effect.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to a crop pest and disease identification method with multi-size input and multi-size targets.
Background
With the rapid development of artificial intelligence technology in the field of computer vision in recent years, the crop pest identification method based on deep learning target detection becomes a research hotspot with the advantages of non-invasion, high speed, low cost and the like. However, in practical application, the pest and disease identification model usually has the problems of multi-size input and multi-size targets, wherein the multi-size input indicates that different shooting cameras are used by users, and the resolutions are different, namely, the input of the pest and disease identification model is multi-size; the multi-size target means that the angle and the distance of the user when the user shoots crops by using the camera are not controllable, so that the size of the target relative to the image can be changed, namely, the target of the pest and disease identification model is multi-size. The two problems cause serious missing detection of the pest and disease identification model when the pest and disease identification model actually falls to the ground, and influence on user experience.
At present, the following methods are available for identifying crop diseases and insect pests: the image classification method is used for classifying the pest and disease damage types of the pictures, cannot position the specific position of a target, is easily influenced by the environment in practical application, and is low in recognition rate especially under the condition that the target is small; based on a general target detection method, rectangular frame positioning is carried out on the target position, the category of the target is identified, the identification effect is easily influenced by a training set, and data of various sizes are required to be supplemented; the method for example segmentation is used for positioning the polygonal area of the target and identifying the target type, and is high in labeling cost and low in speed, and data of various sizes also needs to be supplemented.
Disclosure of Invention
The invention aims to provide a crop pest and disease identification method with multi-size input and multi-size targets, so that the defect of low recognition rate of the existing crop pest and disease is overcome.
In order to achieve the purpose, the invention provides a crop pest and disease identification method with multi-size input and multi-size targets, which comprises the following steps:
collecting an original crop disease and insect pest image, and carrying out target marking on leaves and fruits of diseases and insect pests in the original crop disease and insect pest image;
and converting the marked crop disease and pest image into a data set, wherein the data set comprises the following components: a training set, a verification set and a test set;
constructing a target detection network;
training and verifying the target detection network through the training set and the verification set;
and performing model reasoning on the target detection network after the test set input training and verification to obtain a first reasoning result, and performing sliding window recognition, filling recognition or non-maximum suppression processing on the first reasoning result to obtain a final recognition result.
Preferably, the step of reasoning the target detection network after the test set input training and verification to obtain the final recognition result comprises the following steps:
inputting the test pictures in the test set into a trained and verified target detection network for reasoning to obtain a first reasoning result;
calculating the average ratio test _ w _ ratio and test _ h _ ratio of the width and the height of all rectangular frames in the first reasoning result to the width and the height of the corresponding original picture; the average ratio train _ w _ ratio and train _ h _ ratio of the width and the height of the rectangular frame marked by all targets in the training set to the width and the height of the corresponding original picture;
comparing the test _ w _ ratio and the test _ h _ ratio with the train _ w _ ratio and the train _ h _ ratio;
if test _ w _ ratio < train _ w _ ratio and test _ h _ ratio < train _ h _ ratio, performing sliding window identification on the test picture, then merging results of each sliding window identification, performing non-maximum suppression on the merged sliding window, and selecting an optimal matching rectangular frame, wherein the content in the optimal matching rectangular frame is the final identification result of the test picture;
if the test _ w _ ratio > train _ w _ ratio and the test _ h _ ratio > train _ h _ ratio, filling the test picture, inputting the filled test picture into a trained and verified target detection network for identification, and performing non-maximum suppression to obtain a final identification result;
and if the test _ w _ ratio, the test _ h _ ratio and the train _ w _ ratio, and the train _ h _ ratio are not in the conditions of test _ w _ ratio < train _ w _ ratio, and test _ h _ ratio < train _ h _ ratio, test _ w _ ratio > train _ w _ ratio, and test _ h _ ratio > train _ h _ ratio, directly performing non-maximum suppression on the test picture to obtain the final recognition result of the test picture.
Preferably, the parameter calculation of the sliding window in the sliding window identification comprises:
the height win _ h of the sliding window is calculated as shown in equation (1):
in the formula (1), src _ img _ h is the height of the original picture;
the width win _ w of the sliding window is calculated as shown in equation (2):
in the formula (2), src _ img _ w is the width of the original picture;
the step size step _ h of the up-down sliding window is calculated as shown in equation (3):
step_h＝α*win_h (3)
in the formula (3), alpha is the ratio of the step lengths of the upper sliding window and the lower sliding window;
the step size step _ w of the left and right sliding windows is calculated as shown in equation (4):
step_w＝β*win_w (4)
in the formula (4), β is a ratio of the step lengths of the left and right sliding windows.
Preferably, when the test picture is filled, the parameter calculation of filling includes:
the height filling pad _ h is calculated as shown in equation (5):
the width filling pad _ w is calculated as shown in equation (6):
preferably, the padding is zero padding.
Preferably, the target detection network is a YOLO network.
Compared with the prior art, the invention has the following beneficial effects:
the invention provides a crop pest identification method with multi-size input and multi-size targets, which comprises the steps of collecting an original crop pest image, and carrying out target marking on leaves and fruits of pests in the original crop pest image; and converting the marked crop disease and pest image into a data set, wherein the data set is divided into: a training set, a verification set and a test set; constructing a target detection network; training and verifying the target detection network through a training set and a verification set; the method comprises the steps of carrying out model reasoning on a target detection network after test set input training and verification to obtain a first reasoning result, carrying out sliding window recognition, filling recognition or non-maximum suppression processing on the first reasoning result, further detecting leaves and fruits of plant diseases and insect pests in the first reasoning result to obtain a final recognition result, optimizing a model reasoning link, and effectively improving the recall rate and the accuracy rate of the model to a multi-size input and multi-size target detection model under the condition of not increasing the labeling cost so as to improve the crop plant disease and insect pest recognition effect.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only one embodiment of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
FIG. 1 is a flow chart of a method of crop pest identification with multiple size inputs and multiple size targets of the present invention;
FIG. 2 is a flow diagram of model inference according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of sliding window identification according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating fill recognition in an embodiment of the present invention.
Detailed Description
The technical solutions in the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, the method for identifying crop pests with multi-size input and multi-size targets provided by the invention comprises the following steps:
collecting an original crop disease and insect pest image, and carrying out target marking on leaves and fruits of diseases and insect pests in the original crop disease and insect pest image;
and converting the marked crop disease and pest image into a data set, wherein the data set is divided into: a training set, a verification set and a test set;
constructing a target detection network;
training and verifying the target detection network through a training set and a verification set;
and performing model reasoning on the target detection network after the test set is input into training and verification to obtain a first reasoning result, and performing sliding window recognition, filling recognition or non-maximum suppression processing on the first reasoning result, thereby further detecting the leaves and fruits of the plant diseases and insect pests in the first reasoning result to obtain a final recognition result.
The method comprises the following steps of:
inputting the test pictures in the test set into the trained and verified target detection network for reasoning to obtain a first reasoning result;
calculating the average ratio test _ w _ ratio and test _ h _ ratio of the width of all rectangular frames in the first reasoning result to the width of the corresponding original picture; the average ratio train _ w _ ratio and train _ h _ ratio of the width and the height of the rectangular frame marked by all targets in the training set to the width and the height of the corresponding original picture;
comparing the test _ w _ ratio and the test _ h _ ratio with the train _ w _ ratio and the train _ h _ ratio;
if test _ w _ ratio < train _ w _ ratio and test _ h _ ratio < train _ h _ ratio, performing sliding window identification on the test picture, combining results of each sliding window identification, performing non-maximum suppression operation on the combined sliding window, and selecting a best matching rectangular frame, wherein the content in the best matching rectangular frame is the final identification result of the test picture, namely the image in the best matching rectangular frame in the test picture is the blade and the fruit of the disease and insect pest. The sliding window identification is to adjust the aspect ratio of the target relative to the image to the aspect ratio of the training set aiming at the input images with different sizes in a sliding window mode on the original image, so as to improve the detection effect of the small-size target, and finally improve the overall recall rate and the accuracy rate of the model by inhibiting the operation of the non-maximum value of the sliding window identification result;
and if the test _ w _ ratio > train _ w _ ratio and the test _ h _ ratio > train _ h _ ratio, filling the test picture, inputting the filled test picture into a trained and verified target detection network for identification, and performing non-maximum suppression operation to obtain a final identification result. Filling and identifying, namely adjusting the aspect ratio of the target relative to the image to be the aspect average ratio of the training set aiming at the input images with different sizes in a mode of filling the original image, and improving the detection effect of the large-size target so as to improve the overall recall rate and the accuracy rate of the model;
and if the test _ w _ ratio, the test _ h _ ratio and the train _ w _ ratio, and the train _ h _ ratio are not in the conditions of test _ w _ ratio < train _ w _ ratio, and test _ h _ ratio < train _ h _ ratio, test _ w _ ratio > train _ w _ ratio, and test _ h _ ratio > train _ h _ ratio, directly performing non-maximum suppression operation on the test picture to obtain the final identification result of the test picture.
In sliding window identification, the parameter calculation of the sliding window comprises: height of sliding window win _ h
The height win _ h of the sliding window is calculated as shown in equation (1):
in the formula (1), src _ img _ h is the height of the original picture;
the width win _ w of the sliding window is calculated as shown in equation (2):
in the formula (2), src _ img _ w is the width of the original picture;
the step size step _ h of the up-down sliding window is calculated as shown in equation (3):
step_h＝α*win_h (3)
in the formula (3), alpha is the ratio of the step lengths of the upper sliding window and the lower sliding window;
the step size step _ w of the left and right sliding windows is calculated as shown in equation (4):
step_w＝β*win_w (4)
in the formula (4), β is a ratio of the step lengths of the left and right sliding windows.
Filling the test pictures by adopting zero filling; the parameter calculation for the padding includes:
the height filling pad _ h is calculated as shown in equation (5):
the width filling pad _ w is calculated as shown in equation (6):
one embodiment of the method for identifying crop pests with multiple-size input and multiple-size targets of the present invention is described in detail to enable those skilled in the art to understand the present invention more:
s1, collecting original crop disease and insect pest images, and marking the disease and insect pest blades and fruits of the crop disease and insect pest images with rectangular frames by using a labelme tool;
s2, converting the data subjected to rectangular frame labeling in the step S1 into a target detection data set in a COCO format, randomly dividing the data into a training set, a verification set and a test set according to the proportion of 0.8:0.1:0.1, and counting the average ratio train _ w _ ratio and train _ h _ ratio of the width and the height of all target rectangular frames in the training set relative to the width and the height of the corresponding original picture;
s3, adopting a YOLOv5 as a basic target detection network, zooming the picture of the training set in the step S2 to 608x608x3 as the input of the target detection network, performing feature extraction through the YOLOv5 to obtain 3 output layers with different sizes, wherein the smaller the size of the output layer is, the larger the object responsible for detection is, the loss is calculated through a loss function on the three output layers respectively, loss summation is performed, and the final loss value is subjected to back propagation to update the model parameters;
s4, inputting the training set in the step S2 into the target detection network obtained in the step S3 for training, and inputting the trained target detection network through the verification set in the step S2 for verification;
s5, inputting the test set in step S2 into the target detection network obtained in step S4, and performing model inference to obtain a final recognition result, where fig. 2 shows a flowchart of the model inference provided in the preferred embodiment of the present application, and specifically includes the following steps:
s51, inputting the test set in the step S2 into the target detection network obtained in the step S4 to carry out model reasoning, and after a first reasoning result is obtained, respectively calculating the average ratio test _ w _ ratio and test _ h _ ratio of the width and the height of all rectangular frames relative to the width and the height of the original picture in the first reasoning result;
s52, comparing the test _ w _ ratio and the test _ h _ ratio with the train _ w _ ratio and the train _ h _ ratio; if test _ w _ ratio < train _ w _ ratio and test _ h _ ratio < train _ h _ ratio, then as shown in fig. 3, the sliding window parameters are calculated by the following equations (1) to (4): the method comprises the following steps of (1) sliding window height win _ h, sliding window width win _ w, up-down sliding window step length step _ h and left-right sliding window step length step _ w, wherein the interval between the up-down sliding window step length alpha and the left-right sliding window step length beta is 0-1, a certain overlapping rate between windows is kept, the overlapping rate is usually set to be 0.5, sliding window identification is carried out on a test picture after sliding window parameters are calculated, finally, the identification results of each sliding window are combined, non-maximum value suppression operation is carried out, and the best matching rectangular frame is selected to obtain the final identification result of the test picture; otherwise, go to step S53;
s53, if test _ w _ ratio > train _ w _ ratio and test _ h _ ratio > train _ h _ ratio, then as shown in fig. 4, the filling parameters are calculated according to the following formulas (5) to (6): the height filling pad _ h and the width filling pad _ w are used for performing zero filling on the original image after the filling parameters are calculated, inputting the filled image into the target detection network obtained in the step S4 for identification, and performing non-maximum value suppression operation to obtain a final identification result; otherwise, go to step S54;
and S54, directly carrying out non-maximum suppression operation on the first inference result, and outputting a final recognition result.
Through optimizing the model reasoning link, the recall rate and the accuracy rate of the model to the multi-size input and multi-size target detection model can be effectively improved under the condition of not increasing the marking cost, and then the crop pest and disease identification effect is improved.
The above disclosure is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of changes or modifications within the technical scope of the present invention, and shall be covered by the scope of the present invention.
Claims (6)
1. A crop pest and disease identification method with multi-size input and multi-size targets is characterized in that: the method comprises the following steps:
collecting an original crop disease and insect pest image, and carrying out target marking on leaves and fruits of diseases and insect pests in the original crop disease and insect pest image;
and converting the marked crop disease and pest image into a data set, wherein the data set comprises the following components: a training set, a verification set and a test set;
constructing a target detection network;
training and verifying the target detection network through the training set and the verification set;
and performing model reasoning on the target detection network after the test set input training and verification to obtain a first reasoning result, and performing sliding window recognition, filling recognition or non-maximum suppression processing on the first reasoning result to obtain a final recognition result.
2. A method of crop pest identification with multiple size inputs and multiple size targets as claimed in claim 1, wherein: the step of reasoning the target detection network after the test set input training and verification to obtain the final recognition result comprises the following steps:
inputting the test pictures in the test set into a trained and verified target detection network for reasoning to obtain a first reasoning result;
calculating the average ratio test _ w _ ratio and test _ h _ ratio of the width and the height of all rectangular frames in the first reasoning result to the width and the height of the corresponding original picture; the average ratio train _ w _ ratio and train _ h _ ratio of the width and the height of the rectangular frame marked by all targets in the training set to the width and the height of the corresponding original picture;
comparing the test _ w _ ratio and the test _ h _ ratio with the train _ w _ ratio and the train _ h _ ratio;
if test _ w _ ratio < train _ w _ ratio and test _ h _ ratio < train _ h _ ratio, performing sliding window identification on the test picture, then merging results of each sliding window identification, performing non-maximum suppression on the merged sliding window, and selecting an optimal matching rectangular frame, wherein the content in the optimal matching rectangular frame is the final identification result of the test picture;
if the test _ w _ ratio > train _ w _ ratio and the test _ h _ ratio > train _ h _ ratio, filling the test picture, inputting the filled test picture into a trained and verified target detection network for identification, and performing non-maximum suppression to obtain a final identification result;
and if the test _ w _ ratio, the test _ h _ ratio and the train _ w _ ratio, and the train _ h _ ratio are not in the conditions of test _ w _ ratio < train _ w _ ratio, and test _ h _ ratio < train _ h _ ratio, test _ w _ ratio > train _ w _ ratio, and test _ h _ ratio > train _ h _ ratio, directly performing non-maximum suppression on the test picture to obtain the final recognition result of the test picture.
3. A method of crop pest identification with multiple size inputs and multiple size targets according to claim 2, characterised in that: in the sliding window identification, the parameter calculation of the sliding window comprises the following steps:
the height win _ h of the sliding window is calculated as shown in equation (1):
in the formula (1), src _ img _ h is the height of the original picture;
the width win _ w of the sliding window is calculated as shown in equation (2):
in the formula (2), src _ img _ w is the width of the original picture;
the step size step _ h of the up-down sliding window is calculated as shown in equation (3):
step_h＝α*win_h (3)
in the formula (3), alpha is the ratio of the step lengths of the upper sliding window and the lower sliding window;
the step size step _ w of the left and right sliding windows is calculated as shown in equation (4):
step_w＝β*win_w (4)
in the formula (4), β is a ratio of the step lengths of the left and right sliding windows.
4. A method of crop pest identification with multiple size inputs and multiple size targets according to claim 2, characterised in that: when the test picture is filled, the parameter calculation of filling comprises the following steps:
the height filling pad _ h is calculated as shown in equation (5):
the width filling pad _ w is calculated as shown in equation (6):
5. a method of crop pest identification with multiple size inputs and multiple size targets according to claim 2, characterised in that: the padding is zero padding.
6. A method of crop pest identification with multiple size inputs and multiple size targets as claimed in claim 1, wherein: the target detection network adopts a YOLO network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011606979.9A CN112686152B (en) | 2020-12-30 | 2020-12-30 | Crop pest identification method with multi-size input and multi-size targets |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011606979.9A CN112686152B (en) | 2020-12-30 | 2020-12-30 | Crop pest identification method with multi-size input and multi-size targets |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112686152A true CN112686152A (en) | 2021-04-20 |
| CN112686152B CN112686152B (en) | 2023-06-09 |
Family
ID=75454817
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011606979.9A Active CN112686152B (en) | 2020-12-30 | 2020-12-30 | Crop pest identification method with multi-size input and multi-size targets |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112686152B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114005029A (en) * | 2021-10-20 | 2022-02-01 | 华南农业大学 | Improved yolov5 network-based fingered citron pest and disease identification method and system |
| CN114677553A (en) * | 2021-12-31 | 2022-06-28 | 广西慧云信息技术有限公司 | Image recognition method for solving unbalanced problem of crop disease and insect pest samples |
| CN114842300A (en) * | 2022-05-11 | 2022-08-02 | 安徽农业大学 | Crop disease and insect pest detection method suitable for rainy environment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107818302A (en) * | 2017-10-20 | 2018-03-20 | 中国科学院光电技术研究所 | Non-rigid multi-scale object detection method based on convolutional neural network |
| CN110120036A (en) * | 2019-04-17 | 2019-08-13 | 杭州数据点金科技有限公司 | A kind of multiple dimensioned tire X-ray defect detection method |
| CN110781839A (en) * | 2019-10-29 | 2020-02-11 | 北京环境特性研究所 | Sliding window-based small and medium target identification method in large-size image |
| CN111797760A (en) * | 2020-07-02 | 2020-10-20 | 绵阳职业技术学院 | Improved crop pest and disease identification method based on Retianet |
-
2020
- 2020-12-30 CN CN202011606979.9A patent/CN112686152B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107818302A (en) * | 2017-10-20 | 2018-03-20 | 中国科学院光电技术研究所 | Non-rigid multi-scale object detection method based on convolutional neural network |
| CN110120036A (en) * | 2019-04-17 | 2019-08-13 | 杭州数据点金科技有限公司 | A kind of multiple dimensioned tire X-ray defect detection method |
| CN110781839A (en) * | 2019-10-29 | 2020-02-11 | 北京环境特性研究所 | Sliding window-based small and medium target identification method in large-size image |
| CN111797760A (en) * | 2020-07-02 | 2020-10-20 | 绵阳职业技术学院 | Improved crop pest and disease identification method based on Retianet |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114005029A (en) * | 2021-10-20 | 2022-02-01 | 华南农业大学 | Improved yolov5 network-based fingered citron pest and disease identification method and system |
| CN114005029B (en) * | 2021-10-20 | 2024-04-23 | 华南农业大学 | Method and system for identifying disease and insect pests of bergamot based on improved yolov network |
| CN114677553A (en) * | 2021-12-31 | 2022-06-28 | 广西慧云信息技术有限公司 | Image recognition method for solving unbalanced problem of crop disease and insect pest samples |
| CN114677553B (en) * | 2021-12-31 | 2024-05-14 | 广西慧云信息技术有限公司 | Image recognition method for solving imbalance problem of crop disease and pest samples |
| CN114842300A (en) * | 2022-05-11 | 2022-08-02 | 安徽农业大学 | Crop disease and insect pest detection method suitable for rainy environment |
| CN114842300B (en) * | 2022-05-11 | 2024-05-03 | 安徽农业大学 | Crop pest detection method suitable for rainy day environment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112686152B (en) | 2023-06-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | A computer vision system for early stage grape yield estimation based on shoot detection | |
| CN111723654B (en) | High-altitude parabolic detection method and device based on background modeling, YOLOv3 and self-optimization | |
| CN110909690B (en) | Method for detecting occluded face image based on region generation | |
| CN110532970B (en) | Age and gender attribute analysis method, system, equipment and medium for 2D images of human faces | |
| CN109344883A (en) | Fruit tree diseases and pests recognition methods under a kind of complex background based on empty convolution | |
| CN112686152A (en) | Crop pest and disease identification method with multi-size input and multi-size targets | |
| CN113392775A (en) | Sugarcane seedling automatic identification and counting method based on deep neural network | |
| CN110569747A (en) | method for rapidly counting rice ears of paddy field rice by using image pyramid and fast-RCNN | |
| CN115661943B (en) | Fall detection method based on lightweight attitude assessment network | |
| CN109492665A (en) | Detection method, device and the electronic equipment of growth period duration of rice | |
| CN113298023B (en) | Insect dynamic behavior identification method based on deep learning and image technology | |
| CN113159300A (en) | Image detection neural network model, training method thereof and image detection method | |
| CN107563299B (en) | Pedestrian detection method using RecNN to fuse context information | |
| CN103729620B (en) | A kind of multi-view pedestrian detection method based on multi-view Bayesian network | |
| CN113657294A (en) | Crop disease and insect pest detection method and system based on computer vision | |
| CN114202643A (en) | Apple leaf disease identification terminal and method based on multi-sensor fusion | |
| CN111340019A (en) | Grain bin pest detection method based on Faster R-CNN | |
| CN114663769B (en) | Fruit identification method based on YOLO v5 | |
| CN114758132B (en) | Fruit tree disease and pest identification method and system based on convolutional neural network | |
| CN111339950B (en) | Remote sensing image target detection method | |
| CN113591610A (en) | Crop leaf aphid detection method based on computer vision | |
| CN118521958A (en) | Safety belt detection method and terminal | |
| CN115249357A (en) | Bagged citrus detection method based on semi-supervised SPM-YOLOv5 | |
| CN116612178A (en) | Method for extracting grain shape of integrated rice spike grains | |
| CN116189175A (en) | Crop disease and pest classification and identification method based on InheretofectNet algorithm |
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 |