DE102020215877A1 - Method of classifying plants for agricultural purposes - Google Patents

Abstract

A method is proposed for classifying plant parts (34) in image information, in particular for agricultural purposes, with the following steps:
- identifying plant parts (34) in the image information using a first segmentation threshold value by means of a classification unit in order to obtain plant image information (32);
- Identification of reference plant parts in the image information using a second segmentation threshold value by means of the classification unit in order to obtain reference plant image information, the second segmentation threshold value being different from the first segmentation threshold value and also selected such that the identified reference plant parts of the reference plant image information item contain the identified plant parts (34) of the plant image information (32);
- comparing the identified plant parts (34) with the identified reference plant parts by means of the classification unit; and
- Classifying identified plant parts (34) of the plant image information (32) into the same plant class if they are arranged in a coherent identified reference plant part by means of the classification unit.

Description

State of the art

The invention relates to a method for classifying plant parts in image information, in particular for agricultural purposes, a corresponding classification unit and an agricultural spraying device for applying a medium, in particular liquid spraying agent, according to the species of the independent claims.

The normalized differentiated vegetation index, "NDVI" for short (Normalized Difference Vegetation Index), is an index that shows the biomass, i.e. the proportion of plants, in one image. Plants reflect relatively little radiation in the red spectral range and relatively much radiation in the near infrared spectral range. Infrared cameras or cameras with Bayer sensors can be used to determine the NDVI when observing vegetation.

Here, the leaves of a plant are often better segmented than the rest of the plant. In order to determine the object affiliation of each found object, however, the entire plant must be segmented.

A method for classifying plants is from DE 10 2017 210 804 A1 known. In the known method, an area of a usable area is recorded by means of an optical and/or infrared detection unit and the recorded image data is evaluated by the plants being segmented using threshold values in the NDVI. From this segmentation, the culture series are recognized. The aim is to recognize useful plants that are arranged in the form of plant rows, in order to be able to distinguish between the useful plants and plants located between the useful plants or plant rows, which are classified in particular as weeds, in order to be able to treat them, for example, using appropriate sprays .

Disclosure of Invention

• - Identifizieren von Pflanzenteilen in der Bildinformation unter Verwendung eines ersten Segmentierungsschwellenwerts mittels einer Klassifizierungseinheit, um eine Pflanzenbildinformation zu erhalten;
• - Identifizieren von Referenzpflanzenteilen in der Bildinformation unter Verwendung eines zweiten Segmentierungsschwellenwerts mittels der Klassifizierungseinheit, um eine Referenzpflanzenbildinformation zu erhalten, wobei der zweite Segmentierungsschwellenwert von dem ersten Segmentierungsschwellenwert verschieden und ferner derart gewählt ist, dass die identifizierten Referenzpflanzenteile der Referenzpflanzenbildinformation die identifizierten Pflanzenteile der Pflanzenbildinformation umfasst;
• - Vergleichen der identifizierten Pflanzenteile mit den identifizierten Referenzpflanzenteilen mittels der Klassifizierungseinheit; und
• - Klassifizieren von identifizierten Pflanzenteilen der Pflanzenbildinformation in die gleiche Pflanzenklasse, wenn diese in einem zusammenhängenden identifizierten Referenzpflanzenteil angeordnet sind, mittels der Klassifizierungseinheit.

The subject matter of the present invention is a method for classifying plant parts in image information, in particular for agricultural purposes, with the following steps:

• - identifying parts of plants in the image information using a first segmentation threshold value by means of a classification unit in order to obtain plant image information;
• - Identifying reference plant parts in the image information using a second segmentation threshold value by means of the classification unit in order to obtain reference plant image information, the second segmentation threshold value being different from the first segmentation threshold value and also selected in such a way that the identified reference plant parts of the reference plant image information comprise the identified plant parts of the plant image information;
• - comparing the identified plant parts with the identified reference plant parts by means of the classification unit; and
• - Classifying identified plant parts of the plant image information into the same plant class if they are arranged in a coherent identified reference plant part by means of the classification unit.

• - Empfangen einer Bildinformation mit Pflanzenteilen;
• - Identifizieren von Pflanzenteilen in der Bildinformation unter Verwendung eines ersten Segmentierungsschwellenwerts, um eine Pflanzenbildinformation zu erhalten;
• - Identifizieren von Referenzpflanzenteilen in der Bildinformation unter Verwendung eines zweiten Segmentierungsschwellenwerts, um eine Referenzpflanzenbildinformation zu erhalten, wobei der zweite Segmentierungsschwellenwert von dem ersten Segmentierungsschwellenwert verschieden und ferner derart gewählt ist, dass die identifizierten Referenzpflanzenteile der Referenzpflanzenbildinformation die identifizierten Pflanzenteile der Pflanzenbildinformation umfasst;
• - Vergleichen der identifizierten Pflanzenteile mit den identifizierten Referenzpflanzenteilen; und
• - Klassifizieren von identifizierten Pflanzenteilen der Pflanzenbildinformation in die gleiche Pflanzenklasse, wenn diese in einem zusammenhängenden identifizierten Referenzpflanzenteil angeordnet sind.

The subject matter of the present invention is also a classification unit that is set up to carry out the following steps:

• - Receiving an image information with parts of plants;
• - identifying plant parts in the image information using a first segmentation threshold to obtain plant image information;
• - Identifying reference plant parts in the image information using a second segmentation threshold value to obtain reference plant image information, the second segmentation threshold value being different from the first segmentation threshold value and further selected such that the identified reference plant parts of the reference plant image information comprise the identified plant parts of the plant image information;
• - comparing the identified plant parts with the identified reference plant parts; and
• - Classifying identified plant parts of the plant image information into the same plant class if they are arranged in a contiguous identified reference plant part.

The subject of the present invention is also an agricultural spraying device for applying a medium, in particular a liquid spray, to parts of plants on an agricultural area, with at least one spray nozzle unit for applying the medium, at least one optical detection unit for detecting the parts of plants and a device as described above common classification unit for classifying the plant parts.

Finally, the subject matter of the present invention is a computer program that is set up to execute the steps of a previously described method and/or a previously described classification unit when the computer program is executed on a computer, and a machine-readable storage medium with the computer program stored thereon.

The method or the method steps are, of course, carried out repeatedly.

The method is intended for, but not limited to, agricultural purposes. In the context of the present invention, an agricultural purpose can be understood to mean a purpose which is aimed at the economic cultivation of useful plants.

In the context of the present application, the term “plant parts” includes both individual parts or components of a plant such as plant stalks, plant stalks and plant leaves, but also groups of such individual parts and entire plants.

The step of identifying plant parts can include detecting a color component, in particular an infrared component and/or a green color component in the image information. Red and infrared light are preferably used. Plants or parts of plants reflect relatively little radiation in the red spectral range and relatively much radiation in the near infrared spectral range. Plant parts can be detected by means of the optical detection unit, e.g. using a predetermined NDVI threshold value (Normalized Differenced Vegetation Index, it is formed from reflection values in the near infrared and visible red wavelength range of the light spectrum), by removing biomass or vital plants and plant parts from the ground is/are differentiated. From the normalized difference from both spectral ranges, an NDVI plant image is created, which depicts the parts of the plant and in which the image information is divided into the classes "plant" and "no plant". Identifying parts of plants can, for example, according to the as yet unpublished application DE 10 2020 215 415 take place.

The first segmentation threshold is chosen such that only real plants are segmented. The second segmentation threshold is different from the first segmentation threshold and is further chosen such that the identified reference plant parts of the reference plant image information includes the identified plant parts of the plant image information. In other words, the second segmentation threshold value is selected in such a way that the plant parts identified with the first segmentation threshold value and even larger plant parts that contain them and possibly other plant parts are identified. The second segmentation threshold is preferably chosen such that entire plant stalks and/or entire plant stalks and/or entire plant leaves are identified.

The first segmentation threshold value and/or the second segmentation threshold value is/are preferably selected as a function of a type and/or culture of the plant parts. Alternatively or additionally, the first segmentation threshold value and/or the second segmentation threshold value is/are preferably selected as a function of the optical detection unit.

The segmentation threshold values can be manually enterable, previously determined and transmitted to the system, or preset values and values fixed on the system.

It should be noted here that the choice of the corresponding segmentation threshold value is in particular a level of the value itself. However, it may also be a type of value, without departing from the scope of the present application.

The first segmentation threshold is preferably a first gray value threshold for an image or image information with spectral information suitable for plant segmentation, in particular for a multispectral image with red and near-infrared color information for NDVI calculation, more particularly a first NDVI threshold. Alternatively or additionally, the second segmentation threshold is preferably a second gray value threshold for an image or image information with spectral information suitable for plant segmentation, in particular for a multispectral image with red and near-infrared color information for NDVI calculation, more particularly a second NDVI threshold. Accordingly, the segmentation thresholds can also be green thresholds.

Therefore, more preferably, the second segmentation threshold is lower than the first segmentation threshold. Here, the NDVI threshold values are preferably parameterized depending on the crop type and the optical detection unit or camera system used. Because the second NDVI threshold is lower, weaker reflecting areas in the image information are also identified as "plant" so that plant stalks, plant stalks and complete leaf shapes are found.

Accordingly, the image information preferably has spectral information suitable for plant segmentation. In other words, the image information is an image with spectral information suitable for plant segmentation. In this case, the image information is preferably a multispectral image with red and near-infrared color information for the NDVI calculation, more preferably an NDVI plant image. Accordingly, the image information can also be a monochromatic image with signals in the green wavelength spectrum.

The image information can, for example, represent a field section detected by means of an optical detection unit. The image information is preferably an image or image of the detected field section. Plants are identified and classified from the image information.

A step of capturing a field section with plants by means of an optical capturing unit, in particular an agricultural spraying device, is preferably provided in order to obtain the image information. The step of detecting the field sections can take place, for example, while the agricultural spraying device is traveling over the field with the optical detection unit. Understandably, field sections that are arranged next to one another are hereby detected essentially simultaneously by a plurality of optical detection units. The steps of detecting the field sections in the direction of travel are preferably carried out or carried out repeatedly in a defined, in particular fixed time interval or in a time interval that is adapted to the speed of the agricultural spraying device. This means, in other words, that the field sections are recorded with a defined or a speed-dependent frame rate or frame frequency (=repetition rate).

The optical detection unit is preferably arranged on the agricultural spraying device. The optical detection unit can include at least one multispectral and/or hyperspectral and/or infrared and/or camera and/or 3D camera. The optical detection unit can be designed to detect or record images in the NIR and/or visual range. The optical detection unit can have a light or lighting unit. A plurality of optical detection units can be provided which are designed to communicate with one another. The optical detection unit or the optical detection units is/are preferably arranged on a spray boom of the agricultural spray device.

The plant image information represents image information segmented into “plants” and “non-plants”, with the segmentation result being recorded in the form of objects. This describes very well the shapes and sizes of the plant parts.

Analogously to the plant image information, the reference plant image information also represents image information segmented into “plants” and “non-plants”, with the segmentation result being recorded in the form of objects. Compared to the plant image information, however, the segmentation result of the reference plant image information is more extensive, i.e. it contains the objects of the plant image information in the form of larger objects in which they are contained and possibly other objects that are not contained in the plant image information. Accordingly, the reference plant image information is "blurred" than the plant image information.

In the step of comparing the identified plant parts with the identified reference plant parts, groups of plant parts are determined which are arranged in a coherent identified reference plant part. That is, in other words, that a kind of superimposition of the plant image information with the reference plant image information takes place. As a result, parts of plants which are spatially separated in the image information can be classified as an object group or a plant object or belonging to the same plant.

Consequently, in the classification step, identified plant parts of the plant image information are classified into the same plant class if they are arranged in a coherent identified reference plant part. In this case, the plant parts are preferably assigned to the same plant. Here, the plant parts can be classified into a first plant class of useful plants and/or crop plants or into a second plant class of weeds.

The classification unit includes an algorithm for classifying plants in the image information. The classification unit can be an arithmetic unit or a large number of arithmetic units for processing signals or data, at least one memory unit for storing signals or data, at least one communication interface for reading in data, in particular for receiving image information, and a control unit for outputting data, in particular control signals to a unit, esp which include an actuator. A classification unit can be assigned to each optical detection unit or each optical detection unit can have its own classification unit. The classification unit is or the classification units are designed or set up for image processing, so that it can carry out calculation steps or image processing steps for carrying out the method according to the invention. Accordingly, each classification unit has corresponding image processing software. The arithmetic unit can be a signal processor, a microcontroller or the like, for example, while the storage unit can be a flash memory, an EPROM or a magnetic storage unit. The communication interface can be designed to read in or output data wirelessly and/or by wire, with a communication interface that can read in or output wire-bound data reading this data, for example electrically or optically, from a corresponding data transmission line or outputting it into a corresponding data transmission line.

Accordingly, the method according to the invention can be implemented, for example, in software or hardware or in a mixed form of software and hardware in the classification unit or a control unit.

The classification unit can be arranged completely or partially on or integrated into the agricultural spraying device. However, the classification unit can also be completely or partially external, for example integrated in a cloud.

The method according to the invention enables efficient classification with a high classification quality, in particular of parts of plants that are spatially remote in an image information but belonging to the same plant. Accordingly, plant parts in particular that still belong to the cultivated plant but are spatially separated from their plant due to thin and/or weakly reflecting plant parts, such as plant stalks and plant stalks, can be classified much more precisely during segmentation using the image data.

By generating the reference image information with the second segmentation threshold value or lower NDVI threshold value and the subsequent comparison with the reference image information, these spatially distant plant parts, which were not identified as a plant object in the "sharper" plant image information, can be brought together and considered to belong to the same plant classified or classified together in the correct class.

This is the case, for example, with classification methods or classification algorithms - such as the DE 10 2017 210 804 A1 known - relevant, for which all identified objects or parts of plants are classified as "weeds" by definition if they are not segmented as a coherent object/contiguous plant and exceed a defined minimum distance to an identified crop row and/or do not touch a defined crop row hose

With the method according to the invention, all plant parts are “merged” into one plant object each, as long as there is sufficient reflective plant material between these plant parts, which is identified using the second segmentation threshold value and stored in the reference plant image information. As soon as there is a spot of e.g. earth between parts of plants, these objects are not merged but remain separate. This also ensures that a weed that grows close to the cultivated plant cannot be fused with the cultivated plant and thus continues to exist separately as a weed.

The higher classification quality in turn leads to a more precise application of, for example, plant protection products, since an application may be required due to a correct classification of the plants. Furthermore, the method according to the invention enables a fast and efficient working speed, in particular when treating an agricultural area, for example using a field sprayer, due to the low required runtimes of the necessary image processing operators and computing power.

It is advantageous if the step of identifying reference plant parts comprises a step of dilating the identified reference plant parts in order to obtain the reference plant image information. Here, the segmentation result of the second segmentation threshold value is dilated, i.e. the identified objects or plant parts are expanded by a parameterizable size x using an image processing operator. In this way, "gaps" in and between the identified plant parts can be closed in a simple manner, which can further increase the classification quality.

It is further advantageous if a step of identifying rows of cultivated plants in the plant image information using identified plants by means of the classification unit is provided, wherein in the step of classifying the plant parts are also classified as a function of position information, in particular a distance of the plant parts to the corresponding crop row.

The step of identifying or recognizing crop rows can be carried out continuously or permanently. The step of identifying crop rows is preferably carried out using and/or evaluating all or essentially all of the image information obtained.

Advantageously, the row of cultivated plants can be identified using at least one of the following items of information: color component, in particular green color component of the detected plants, infrared component of the detected plants, distance between plants, distance between rows of plants, growth stage of the plants, geo-coordinates of a sowing of the plants, direction of movement of the optical detection unit or direction of travel the agricultural sprayer. Using this information or properties, the rows of crop plants can be identified in a simple and reliable manner. Since rows of cultivated plants essentially extend in a straight line, a row of cultivated plants can also be identified, for example, by fitting a straight line or a straight center line of cultivated plant rows into an image trajectory with the highest corresponding color component or color value. In the identification step, all rows of cultivated plants are preferably identified in the image information.

A step of applying a medium, in particular a liquid spray, to the classified plant parts of a corresponding plant class, in particular to the classified plant parts in the image evaluation area, is preferably provided as a function of the determined plant index by means of a spray nozzle unit, in particular an agricultural spray device.

In this case, the medium or spray agent is preferably applied to an agricultural area or an area used for agriculture. This can be understood to mean a field or an area under cultivation for plants or also a parcel of such an area under cultivation. The agricultural area can thus be arable land, grassland or pasture. The plants can include, for example, useful plants whose fruit is used agriculturally (for example as food, animal feed or as an energy crop) and weeds, weeds and grass weeds.

All steps of the method are preferably carried out during a movement, in particular a drive or a flight of the agricultural spraying device over the agricultural area. Thus, within the scope of the present application, depending on the configuration of the spraying device, the term “direction of travel” can also be understood as a “direction of flight”.

The agricultural spraying device is advantageously designed to carry out the method automatically and/or autonomously in order to enable a field to be treated quickly, reliably and efficiently.

The agricultural spraying device can in particular be part of an agricultural field sprayer or a crop protection device or be designed as an agricultural field sprayer or a crop protection device. The agricultural spraying device can comprise a mobile unit or be arranged on a mobile unit, wherein the mobile unit can be designed in particular as a land vehicle and/or aircraft and/or trailer. The mobile unit can in particular be an agricultural working machine, for example a tractor, a tractor, a self-propelled or autonomous field sprayer or a self-propelled or autonomous robot. The agricultural spraying device can in particular be a towed field sprayer, a self-propelled field sprayer or a mounted field sprayer. The agricultural spray device can also be attached to a hydraulic device of an agricultural working machine. It is also conceivable that the agricultural spraying device is installed on a loading area of an agricultural working machine. Alternatively, the spray device can be attached to the agricultural working machine. The agricultural spray device or the field sprayer can have at least one spray agent tank for receiving the spray agent. The agricultural spray device or the field sprayer can also have a mixing unit which mixes a spray concentrate with water directly on the agricultural spray device to form the spray to be applied.

The medium or spray is in particular a spray liquid. The spray can include or be an agricultural preparation or plant protection agent (PPP), in particular a plant protection agent concentrate. Accordingly, the spray can contain a pesticide, such as a herbicide, fungicide or an insecticide. However, the spray can also include or be a fertilizer, in particular a fertilizer concentrate. The spray can have a growth regulator. The spray can be granular Have active agent, which was mixed with a carrier liquid. The spray liquid can be designed, for example, as: liquid, suspension, emulsion, solution or a combination thereof. The spray liquid is preferably in the form of a plant protection agent diluted with water or a fertilizer diluted with water. The spray liquid can accordingly be a spray mixture, for example.

The dispensing of the medium or spray agent can be carried out in particular by means of a delivery unit. In this case, the delivery unit can be designed to deliver or direct, in particular meter, a liquid and/or granules under pressure. Accordingly, the delivery unit can include, for example, one/one or more pumps, delivery pumps, dosing pumps, pressure accumulators, screw conveyors, valves, diaphragms, etc.

The spray nozzle unit preferably has at least one spray nozzle for discharging the spray agent and at least one valve for controlling or regulating the quantity of spray agent that is discharged. Accordingly, the spray nozzle unit is designed to be controllable or operable. The valve can be arranged or integrated in the spray nozzle. However, the valve can also be arranged upstream of the spray nozzle, i.e. upstream of the spray nozzle (in the direction of flow of the spray). However, the spray nozzle unit can also have several spray nozzles, each with an upstream valve. Furthermore, the spray nozzle unit can also have a plurality of spray nozzles with only one valve connected upstream of the spray nozzles, so that when the valve is actuated, the spray agent is applied by means of all the spray nozzles of the spray nozzle unit. The valve can be designed as a pulse width modulated valve (PWM valve) or as a proportional valve. The spray nozzle unit can be designed as part width of a nozzle system of the agricultural spray device. The spray nozzle units can be controlled individually or separately and/or in defined groups or associations and/or all together. The spray nozzles of each spray nozzle unit can be controlled individually or separately and/or in defined groups or associations and/or all together.

The spray nozzle unit or the spray nozzle units is/are preferably arranged on a spray boom of the agricultural spray device.

A step of determining a plant index for at least one defined image evaluation area in the plant image information using classified plant parts in the respective defined image evaluation area by means of the classification unit is preferably provided. The plant code preferably represents a degree of coverage of the image analysis area by plant material and/or a quantity of plant material in the image analysis area and/or a number of classified plant parts in the image analysis area. The degree of coverage can be defined as the ratio of the area covered by plant material to the total area to be evaluated. The degree of coverage for the image evaluation area is therefore the ratio of the area of the overgrown area to the respective overall image evaluation area. For this purpose, the number of pixels in the respective image evaluation area can be determined at which plant material is detected. Accordingly, a measure of the infestation can be derived by means of the plant index, on the basis of which a decision is made as to whether and, if so, how (e.g. with which application quantity) a corresponding field area is sprayed or treated.

The threshold can be manually entered, a previously determined value and transmitted to the system, or a preset value and fixed value on the system. Depending on the crop in the field, the growth stage and the spray or pesticide used, a so-called spraying rule can be stored, i.e. a connection between a specific plant index and the decision as to whether and how much pesticide should be applied. As an example, the rule could be: "If the degree of coverage in the evaluation area exceeds 0.5%, then it is applied to the field area".

It should be noted that further separate image evaluation areas can be provided in the image information without departing from the scope of the present invention. The number of image evaluation areas per image information depends on the distance between the spray nozzles and the width of the image information. Each image evaluation area can be assigned to a spray nozzle unit with one or more spray nozzles. For example, with an image information width of 1.5 m and a spray nozzle distance of 25 cm, 6 image evaluation areas would be provided per camera.

Thus, the image evaluation areas are the areas of the image information in which the plant parameters are determined and evaluated in order to apply the medium or spray agent depending on this. In this case, an image evaluation area represents a corresponding field area of the recorded field sections, so that in the application step the spray agent is preferably applied to the corresponding field area.

In the application step, if the “spraying decision” is positive, the spray agent is preferred depending on the plant determined zen indicators, in particular when a defined threshold value for the plant indicator is reached and/or fallen below and/or exceeded by means of the spray nozzle unit of the agricultural spray device. The spray is preferably applied with a defined minimum amount per area. In this case, the spray is preferably applied to the entire field area.

character list

• 1 eine schematische Darstellung einer erfindungsgemäßen landwirtschaftlichen Spritzvorrichtung;
• 2 Farbbild mit einer Soja-Pflanze;
• 3 ein NDVI-Pflanzenbild des Farbbilds aus 1;
• 4 eine Pflanzenbildinformation mit mittels des ersten Segmentierungsschwellenwertes identifizierten Pflanzenteilen aus dem NDVI-Pflanzenbild von 3;
• 5 die Pflanzenbildinformation aus 4 mit klassifizierten Pflanzenteilen gemäß dem Stand der Technik;
• 6 eine erfindungsgemäße Referenzpflanzenbildinformation mit mittels des zweiten Segmentierungsschwellenwertes identifizierten Pflanzenteilen aus dem NDVI-Pflanzenbild von 3;
• 7 ein Vergleichsergebnis der Überlagerung der Pflanzenbildinformation aus 4 mit der Referenzpflanzenbildinformation aus 6;
• 8 die Pflanzenbildinformation aus 4 mit klassifizierten Pflanzenteilen gemäß der vorliegenden Erfindung; und
• 9 ein Ablaufdiagramm eines Verfahrens gemäß einem Ausführungsbeispiel.

The invention is explained in more detail below by way of example with reference to the attached drawings. Show it:

• 1 a schematic representation of an agricultural sprayer according to the invention;
• 2 color image with a soybean plant;
• 3 an NDVI plant image of the color image 1 ;
• 4 a plant image information with plant parts identified by means of the first segmentation threshold from the NDVI plant image of 3 ;
• 5 the plant image information 4 with classified plant parts according to the state of the art;
• 6 a piece of reference plant image information according to the invention with plant parts identified by means of the second segmentation threshold value from the NDVI plant image of 3 ;
• 7 a comparison result of the superposition of the plant image information 4 with the reference plant image information 6 ;
• 8th the plant image information 4 with classified plant parts according to the present invention; and
• 9 a flow chart of a method according to an embodiment.

In the following description of preferred exemplary embodiments of the present invention, the same or similar reference symbols are used for the elements which are shown in the various figures and have a similar effect, with a repeated description of the elements being dispensed with.

In 1 1 is a schematic representation of an agricultural sprayer, which is provided with the reference numeral 10 in its entirety.

The agricultural spraying device 10 is designed as a field sprayer 10 . The field sprayer 10 is arranged on a mobile land vehicle 12 which is designed as a towing vehicle 12 or tractor 12 .

The agricultural spray device 10 has a spray boom 14 . Spray nozzle units 16 and optical detection units 18 are arranged on the spray boom 14 . The spray nozzle units 16 are designed to apply a spray 20 to plants 22 or weeds 23 on an agricultural area 24 . The optical detection units 18 are in the form of optical cameras 18 . The optical cameras 18 each include a filter unit in order to extract a color component, such as the green color component, of image information obtained or captured or of a captured image in order to capture crop plants 22 and weeds 23 .

The agricultural spraying device 10 also has a delivery unit (not shown), by means of which the application quantity or an active agent quantity in the spray agent 20 to be applied can be adjusted or varied.

The agricultural sprayer 10 also includes classification units (not shown) connected to the optical cameras 18 to receive information therefrom. In this case, each optical camera 18 is assigned its own classification unit. The classification units have a computing unit which is designed to carry out calculation steps or image processing steps for carrying out the method according to the invention. The classification units are also designed to output a control signal in such a way that the spray agent 20 is applied by means of the associated spray nozzle units 16 as a function of the plant identification number determined.

2 12 shows image information 26 with a soybean plant 22. Image information 26 is an image 26 with spectral information suitable for plant segmentation. Ground 28 of agricultural area 24 can be seen in the background.

In 3 further image information 30 is shown. In this case, the image information 30 is an NDVI plant image 30 of the image 26 1 .

In 4 a piece of plant image information 32 is shown. The plant image information 32 illustrates plant parts 34 identified by means of the first segmentation threshold value or first NDVI threshold value from the NDVI plant image 30 of FIG 3 .

5 shows the plant image information 32 4 with classified plant parts 34 according to the prior art mentioned in the introduction. In this case, some parts of plants 34 were classified in the same plant class, namely as “crop” 22 , due to the arrangement or proximity to a row of crop plants (not shown). In contrast to this, the remaining plant parts 34 were classified in a different plant class, namely as “weeds” 23, due to a lack of relationship or connection to the other plant parts 34 and the distance to the row of cultivated plants.

In 6 a piece of reference plant image information 36 according to the invention is now shown. The reference plant image information 36 illustrates plant parts 38 identified by means of the second segmentation threshold value or second NDVI threshold value from the NDVI plant image 30 of FIG 3 . Here, the second NDVI threshold is lower than the first NDVI threshold of the plant information 32 4 .

How out 6 as can be seen, the choice of the lower second NDVI threshold compared to the plant image information has 32 of 4 led to the fact that other plant parts 38, in the present case including a plant stalk of the soybean plant 22, and larger plant parts 38 were identified. This is made out in particular 7 clearly, in which a comparison result of the superimposition of the plant image information 32 from 4 with the reference plant image information 6 out 6 is shown. For reasons of clarity, the identified plant parts 34 are shown hatched.

Analogous to the classification result according to the prior art 5 shows is now in 8th the plant image information 32 out 4 with classified plant parts 34 according to the present invention. From this it can be seen that according to 5 Plant parts 34 classified in the plant class "Weed" 23 are now correctly also classified in the plant class "Cultivated plant" 22 as these are arranged in a contiguous identified reference plant part 38 with plant parts 34 classified by definition as "Cultivated plant" 22.

9 shows a flowchart of a method 100 for classifying parts of plants 34, 38 in image information 26, 30, in particular for agricultural purposes. The method 100 comprises a step of identifying 104 plant parts 34 in the image information 26, 30 using a first segmentation threshold value by means of a classification unit in order to obtain plant image information 32. The method 100 further comprises a step of identifying 108 reference plant parts 38 in the image information 26, 30 using a second segmentation threshold value by means of the classification unit in order to obtain reference plant image information 36, the second segmentation threshold value being different from the first segmentation threshold value and also being chosen such that that the identified reference plant parts 38 of the reference plant image information 36 includes the identified plant parts 34 of the plant image information 32 . The method 100 also includes a step of comparing 112 the identified plant parts 34 with the identified reference plant parts 38 using the classification unit. The method also includes a step of classifying 114 identified plant parts 34 of the plant image information 32 into the same plant class if they are arranged in a coherent identified reference plant part 38 by means of the classification unit.

The method 100 also includes an optional step of detecting 102 a field section with plants 22, 23 by means of an optical detection unit 18, in particular an agricultural spraying device 10, in order to obtain the image information 26, 30. The method 100 also includes an optional step of identifying 106 rows of crop plants in the plant image information 32 using identified plant parts 34 by means of the classification unit, wherein in the step of classifying 114 the plant parts 34 also as a function of position information, in particular a distance between the plant parts 34 classified according to the relevant crop series. The method 100 further comprises an optional step of identifying 108 reference plant parts 38 comprising a step of dilating 110 the identified reference plant parts 38 in order to obtain the reference plant image information 36 . Finally, the method 100 includes an optional step of determining 116 a plant identifier for at least one defined image evaluation area in the plant image information 32 using classified plant parts 34 in the respective defined image evaluation area by means of the classification unit.

If an embodiment includes an "and/or" link between a first feature and a second feature, this should be read in such a way that the embodiment according to one embodiment includes both the first feature and the second feature and according to a further embodiment either only that having the first feature or only the second feature.

QUOTES INCLUDED IN DESCRIPTION

This list of documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• DE 102017210804 A1 [0004, 0032]
• DE 102020215415 [0012]

Claims (18)

Method (100) for classifying plant parts (34, 38) in image information (26, 30), in particular for agricultural purposes, with the following steps: - identifying (104) plant parts (34) in the image information (26, 30) using a first segmentation threshold value by means of a classification unit in order to obtain plant image information (32); - Identifying (108) reference plant parts (38) in the image information (26, 30) using a second segmentation threshold value by means of the classification unit in order to obtain reference plant image information (36), the second segmentation threshold value being different from the first segmentation threshold value and also being selected in this way that the identified reference plant parts (38) of the reference plant image information (36) comprises the identified plant parts (34) of the plant image information (32); - Comparing (112) the identified plant parts (34) with the identified reference plant parts (38) by means of the classification unit; and - Classifying (114) of identified plant parts (34) of the plant image information (32) into the same plant class if they are arranged in a coherent identified reference plant part (38) by means of the classification unit. Method (100) according to claim 1 , characterized in that the second segmentation threshold is selected such that whole plant stalks (38) and/or whole plant stalks (38) and/or whole plant leaves (38) are identified. Method (100) according to claim 1 or 2 , characterized in that the second segmentation threshold is lower than the first segmentation threshold. Method (100) according to one of the preceding claims, characterized in that - the image information (26, 30) has spectral information suitable for plant segmentation, in particular a multispectral image (30) with red and near-infrared color information for NDVI calculation, more particularly an NDVI - plant image (30); and/or - the first segmentation threshold value is a first gray value threshold value for an image (26, 30) with spectral information suitable for plant segmentation, in particular for a multispectral image (30) with red and near-infrared color information for NDVI calculation, more particularly a first NDVI - Threshold value, and the second segmentation threshold value is a second gray value threshold value for an image (26, 30) with spectral information suitable for plant segmentation, in particular for a multispectral image (30) with red and near-infrared color information for NDVI calculation, more particularly a second NDVI -Threshold are. Method (100) according to one of the preceding claims, characterized in that the first segmentation threshold value and/or the second segmentation threshold value is/are selected as a function of a type and/or culture of the plants (22, 23). Method (100) according to one of the preceding claims, characterized in that the step of identifying (108) reference plant parts (38) comprises a step of dilating (110) the identified reference plant parts (38) in order to obtain the reference plant image information (36). Method (100) according to one of the preceding claims, characterized in that in the step of classifying (114) identified plant parts (34) are assigned to the same plant (22, 23) if they are arranged in a coherent identified reference plant part (38). Method (100) according to one of the preceding claims, characterized in that in the step of classifying (114) identified parts of plants (34) classified in a first plant class of useful plants and / or crops (22) or in a second plant class of weeds (23). will. Method (100) according to one of the preceding claims, characterized by a step of identifying (106) crop rows in the plant image information (32) using identified plant parts (34) by means of the classification unit, wherein in the step of classifying (114) the plant parts ( 34) are also classified as a function of position information, in particular a distance of the plant parts (34) from the corresponding crop row. Method (100) according to one of the preceding claims, characterized by a step of capturing (102) a field section with plants (22, 23) by means of an optical capturing unit (18), in particular an agricultural spraying device (10), in order to display the image information (26, 30) to get. Method (100) according to claim 10 , characterized in that the first segmentation threshold value and/or the second segmentation threshold value is selected as a function of the optical detection unit (18). Method (100) according to one of the preceding claims, characterized by a step of determining (116) a plant identification number for at least one defined image evaluation area in the plant image information (32) using classified plant parts (34) in the respective defined image evaluation area by means of the classification unit. Method (100) according to claim 12 , characterized in that the plant index represents a degree of coverage of the image analysis area by plant material and/or a quantity of plant material in the image analysis area and/or a number of classified plant parts (34) in the image analysis area. Method (100) according to one of the preceding claims, characterized by a step of applying (118) a medium (20), in particular liquid spray (20) to the classified plant parts (34) of a corresponding plant class, in particular to the classified plant parts (34) in the image evaluation area, depending on the determined plant identification number, by means of a spray nozzle unit (16), in particular an agricultural spray device (10). Classification unit that is set up to carry out the following steps: - Receiving an image information (26, 30) with plants (22, 23); - identifying (104) plant parts (34) in the image information (26, 30) using a first segmentation threshold to obtain plant image information (32); - Identifying (108) reference plant parts (38) in the image information (26, 30) using a second segmentation threshold to obtain reference plant image information (36), wherein the second segmentation threshold is different from the first segmentation threshold and is further selected such that the identified reference plant parts (38) of the reference plant image information (36) comprises the identified plant parts (34) of the plant image information (32); - comparing (112) the identified plant parts (34) with the identified reference plant parts (38); and - Classifying (114) of identified plant parts (34) of the plant image information (32) into the same plant class if they are arranged in a contiguous identified reference plant part (38). Agricultural spraying device (10) for applying a medium (20), in particular liquid spray (20) to plant parts (34) on an agricultural area (24), with at least one spray nozzle unit (16) for applying the medium (20), at least one optical detection unit ( 18) for detecting the plant parts (34) and a classification unit for classifying the plant parts (34) according to claim 15 . Computer program that is set up to carry out the steps of a method according to one of Claims 1 until 14 and/or a classification unit claim 15 to be executed when the computer program is run on a computer. Machine-readable storage medium with the computer program stored on it Claim 17 .

Images