DE102021214202A1 - Method for identifying weeds in a row of plants in an agricultural area - Google Patents

Abstract

The invention relates to a method for identifying weeds (18) in a row of plants (20) on an agricultural area, wherein an extent index for defined plant sections (26) of a defined crop plant area is determined in image information (10), the extent index having a ratio of one width value and a length value of the respective defined plant section (26), the image information (10) being divided into transverse regions (28) and longitudinal regions (30), the length value being the number of transverse regions (28) over which the respective defined Plant section (26) extends in a defined longitudinal direction (34), and the width value is a sum of section width values, the section width values indicating the number of longitudinal regions (30) over which the respective defined plant section (26) extends in the respective transverse regions (28) starting from a defined plant row center line (22) of the respective plant row (20) in a defined transverse direction (38), the extent index of the respective plant section with a frequency distribution of extent index of defined plant sections (26) of previously identified plants (16, 18) are compared and the plants (16, 18) which have an associated plant section (26) with an extent index whose frequency is less than or equal to a defined frequency threshold value of the frequency distribution are identified as weeds (18) in an identified plant row (20). .

Description

State of the art

The invention is based on a method for identifying weeds in a row of plants on an agricultural area and a corresponding computing unit, a plant identification unit and an agricultural working machine according to the species of the independent claims. The subject matter of the present invention is also a computer program and a machine-readable storage medium

The DE 10 2017 210 804 A1 discloses a method for applying a spray to a field, the spray being applied depending on the degree of coverage of an evaluation area. When evaluating the image data, plants are segmented using a threshold in the NDVI. From this segmentation, the culture series are recognized. A culture row hose with a certain width is placed around the culture rows. Any segmented object (plant) between the tubes is classified as a weed by definition. All objects that lie within this hose width of the rows or are connected to objects in this hose width are classified as crops by definition.

Other algorithms use networks (deep learning) or classification methods to identify plants. This requires a lot of training data, a lot of label data and offline training in advance. In addition, these methods are very computationally intensive.

Disclosure of Invention

• - Empfangen einer Bildinformation von einem mittels einer optischen Erfassungseinheit erfassten Feldabschnitt einer landwirtschaftlichen Fläche mit Pflanzen;
• - Identifizieren von zumindest einer Pflanzenreihe in der Bildinformation unter Verwendung von identifizierten Pflanzen mittels einer Recheneinheit;
• - Definieren eines die zumindest eine identifizierte Pflanzenreihe umfassenden Kulturpflanzenbereiches unter Verwendung der zumindest einen identifizierten Pflanzenreihe in der Bildinformation mittels der Recheneinheit;
• - Ermitteln jeweils einer Ausdehnungskennzahl für definierte Pflanzenabschnitte der identifizierten Pflanzen des definierten Kulturpflanzenbereiches in der Bildinformation mittels der Recheneinheit, wobei die Ausdehnungskennzahl ein Verhältnis eines Breitenwerts und eines Längenwerts des jeweiligen definierten Pflanzenabschnitts repräsentiert, wobei die Bildinformation in Querbereiche und in Längsbereiche unterteilt ist, wobei der Längenwert die Anzahl der Querbereiche, über die sich der jeweilige definierte Pflanzenabschnitt in einer definierten Längsrichtung erstreckt, und der Breitenwert eine Summe von Teilbreitenwerten ist, wobei die Teilbreitenwerte die Anzahl der Längsbereiche, über die sich der jeweilige definierte Pflanzenabschnitt in den jeweiligen Querbereichen ausgehend von einer definierten Pflanzenreihenmittelpunktlinie der jeweiligen Pflanzenreihe in eine definierte Querrichtung erstreckt, repräsentieren;;
• - Vergleichen der ermittelten Ausdehnungskennzahl mit einer Häufigkeitsverteilung von Ausdehnungskennzahlen von definierten Pflanzenabschnitten vorhergehend identifizierter Pflanzen des definierten Kulturpflanzenbereiches und/oder definierter Kulturpflanzenbereiche vorhergehend erfasster Bildinformationen der landwirtschaftlichen Fläche mittels der Recheneinheit; und
• - Identifizieren der Pflanzenabschnitte mit einer Ausdehnungskennzahl, dessen Häufigkeit kleiner oder gleich einem definierten Häufigkeitsschwellenwert der Häufigkeitsverteilung ist, um die den Pflanzenabschnitten zugehörigen Pflanzen als Beikräuter in einer identifizierten Pflanzenreihe zu identifizieren.

The subject matter of the present invention is a method for identifying weeds, in particular large weeds, in a row of plants on an agricultural area, with the steps:

• - receiving image information from a field section of an agricultural area with plants detected by means of an optical detection unit;
• - identifying at least one row of plants in the image information using identified plants by means of a computing unit;
• - defining a crop plant area comprising the at least one identified row of plants using the at least one identified row of plants in the image information by means of the computing unit;
• - Determination of an extent index for defined plant sections of the identified plants of the defined crop plant area in the image information by means of the computing unit, the extent index representing a ratio of a width value and a length value of the respective defined plant section, the image information being divided into transverse areas and longitudinal areas, the Length value is the number of transverse areas over which the respective defined plant section extends in a defined longitudinal direction, and the width value is a sum of part width values, the part width values being the number of longitudinal areas over which the respective defined plant section extends in the respective transverse areas, starting from a defined plant row center line of the respective plant row extending in a defined transverse direction;
• - Comparing the determined extent index with a frequency distribution of extent indexes of defined plant sections of previously identified plants of the defined crop area and/or defined crop areas of previously recorded image information of the agricultural area by means of the computing unit; and
• - Identifying the plant sections with an extent index whose frequency is less than or equal to a defined frequency threshold value of the frequency distribution in order to identify the plants belonging to the plant sections as weeds in an identified plant row.

The subject matter of the present invention is also a computing unit which is set up to carry out and/or control the steps of a previously described method.

The subject matter of the present invention is also a plant identification unit with an optical detection unit for detecting a field section of an agricultural area with plants in order to obtain image information from the detected field section, and a previously described computing unit.

The subject of the present invention is also an agricultural working machine, in particular a field sprayer, with an agricultural working tool, in particular a spray device, and a previously described plant identification unit, the working tool, in particular the spray device, depending on the identifi adorned weeds in the identified row of plants, is controlled by means of the computing unit.

The present invention also relates to a computer program that is set up to carry out and/or control the steps of a method described above when the computer program is executed on a computer, and a machine-readable storage medium on which the computer program is stored.

The method according to the invention now makes it possible to identify large weeds in a row of plants or rows of cultivated plants in a very simple and resource-saving manner, without requiring an image database or the like. The method can therefore be used very flexibly, as it takes advantage of the extent, i.e. the length and width, of the plants and weeds actually on the agricultural area and extracts them precisely in order to identify large weeds in the plant rows.

According to the invention, this is achieved in that expansion parameters, which represent a ratio between width and length, are determined for defined plant sections of a defined crop area in the image information and are compared with a frequency distribution of expansion parameters of previously identified plant sections. This means, in other words, that a distinction is made in the current crop row between crops and large weeds by means of a statistical evaluation of expansion parameters of plant sections in identified plant rows or within defined crop areas, which is based on a history of previous plants in the plant row.

Here, no ordinary classification method is used, which carries out a clustering from the lengths and then determines which class (crop or weed) an object most likely belongs to. A classifier is a method that ends up calculating a probability of how well an object fits into a class.

Instead, it is determined how often an extent index (=ratio of a width value to a length value) occurs and those objects with the extent index whose frequency is less than or equal to a defined frequency threshold value of the frequency distribution are assumed to be large weeds.

• - Es resultierte eine verbesserte Identifizierung der Beikräuter
• - Es ist keine Sammlung von vielen Bilddaten und kein Anfertigen von Labels notwendig, um ein Netz oder Klassifikation zu trainieren, wodurch der Gesamtaufwand stark reduziert wird.
• - Es ist kein Offline-Training anhand von Bilddaten im Vorfeld notwendig.
• - Die erforderliche Rechenzeit für die Identifizierung oder Klassifizierung anhand eines oder weniger Pflanzenmerkmale ist sehr gering.
• - Die Information über die identifizierten Beikräuter in dem Kulturpflanzenbereich kann für darauffolgende Schritte genutzt werden, um Pflanzenreihen besser bzw. genauer zu identifizieren, da die identifizierten bzw. erkannten Beikräuter hierfür unberücksichtigt bleiben bzw. „ignoriert“ werden können.

This offers a number of advantages:

• - An improved identification of the weeds resulted
• - There is no need to collect a lot of image data or create labels to train a network or classification, which greatly reduces the overall effort.
• - No offline training using image data is necessary in advance.
• - The computing time required for identification or classification based on one or a few plant characteristics is very low.
• - The information about the identified weeds in the crop area can be used for subsequent steps in order to identify rows of plants better or more precisely, since the identified or recognized weeds are not taken into account for this or can be "ignored".

An agricultural area can be understood to mean an area used for agriculture, an area under cultivation for plants or also a parcel of such an area or area under cultivation. The agricultural area can thus be arable land, grassland or pasture. The plants include cultivated plants or useful plants, the fruit of which is used agriculturally, for example as food, animal feed or as an energy crop, and weeds or weeds.

The field section can be a detection section or a detected image section of an optical detection unit. The image information can, for example, be an image of the detected field section. An optical detection unit can be understood to mean, for example, a camera or a 3D camera or an infrared detection unit. The optical detection unit can be calibrated to e.g. B. to calculate the height assignment from captured images.

The method can include a step of capturing a field section of an agricultural area with plants by means of the optical capturing unit. The detecting step can be performed during a transit or a flight of the plant identification unit. At least one further step of the method, in particular all steps of the method, can be carried out during a crossing or a flight of the plant identification unit.

Here, the plant identification unit 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 also be a self-propelled or autonomous robot. The plant identification unit is preferably part of an agricultural working machine. The agricultural working machine is preferably a weed regulating machine, in particular a field sprayer. The agricultural working tool is preferably a spray device, but can also be a mechanical tool for weed control.

The method includes a step of identifying at least one row of plants or a row of cultivated plants in the image information of the recorded field section by means of the computing unit. The at least one row of plants is preferably identified using at least one of the following items of information: color component, in particular red color component of plants in the field section detected, infrared component of plants in the field section detected, distance between plants, distance between rows of plants, growth stage of the plants, geo-coordinates of a sowing of the plants. Using this information, the rows of plants can be identified in a simple manner, since, for example, crop plants are usually planted equidistantly or the crop plants are more advanced in growth stage than the weeds or weeds. In the step of identifying the at least one row of plants, all rows of plants are preferably identified in the image information or in the recorded field section.

The step of identifying at least one row of plants understandably includes the detection of plants or plant parts in the image information or in the detected field section. A detection of plants can be understood, for example, as determining the presence of plants or plant mass/biomass in the field section, in particular without the individual plants being classified in the process. The step of detecting plants can include detecting a color component, in particular a red color component and/or an infrared component in the field section or image section. Here, by means of the optical detection unit, e.g. by evaluating light spectra or using a predetermined NDVI 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 plant mass/biomass from (soil )soil is distinguished, plants are recorded.

The method also includes a step of defining a crop plant area comprising the at least one identified plant row using the at least one identified plant row by means of the computing unit. The crop area is preferably defined using neighborhood pixels of detected plants of the crop area. In this case, the cultivated plant area is preferably defined around a generated plant row center line, in particular with the generated plant row center line extending essentially in a straight line. The crop range may fully encompass the plants of the plant row. However, the cultivated plant area can also include the row of plants, without the individual plants of the row of plants being fully included.

Accordingly, the cultivated plant area can also only partially include the individual plants of the plant row. In other words, all identified plants that are at least partially arranged in the respective defined crop area are assigned to the crop area or are evaluated or viewed as plants of the crop area.

The cultivated plant area can be defined around the respective plant row center line with a constant or defined width. However, the crop area can also have a variable width, wherein the width can depend on a growth stage of a plant arranged in a corresponding area of the crop area. The cultivated plant area is thus designed in the form of a tube. The crop area has a smaller width than the recorded field section or the corresponding image information.

The method also includes a step of determining a respective expansion index for defined plant sections of the identified plants of the defined cultivated plant area in the image information using the computing unit.

The method also includes a step of determining an extent index for defined plant sections of the identified plants of the defined crop plant area in the image information using the computing unit, with the extent index representing a ratio of a latitude value and a longitude value of the respective defined plant section, the image information being divided into transverse areas and is divided into longitudinal areas, where the length value is the number of transverse areas over which the respective defined plant section extends in a defined longitudinal direction, and the width value is a sum of part width values, wherein the partial width values represent the number of longitudinal areas over which the respective defined plant section extends in the respective transverse areas, starting from a defined plant row center line of the respective plant row in a defined transverse direction.

Advantageously, the defined longitudinal direction runs along, i.e. parallel, and the defined transverse direction runs transversely, in particular at an angle of greater than or equal to 85° to less than or equal to 90°, i.e. substantially perpendicular to the plant row center line of the identified plant row. In this case, the transverse regions preferably have essentially the same length and/or the longitudinal regions have essentially the same width.

Thus, the transverse areas run transversely, preferably at an angle of greater than or equal to 85° to less than or equal to 90°, i.e. essentially perpendicular to the defined longitudinal direction or the plant row center line. However, the transverse areas can also run horizontally in the image information or the image independently of the defined longitudinal direction or the plant row center line. The transverse areas preferably have a substantially equal length in the defined longitudinal direction. In this case, the length of the transverse areas is greater than 1 pixel. The length of the transverse regions is preferably in a range from greater than or equal to 5 pixels to less than or equal to 50 pixels.

Similarly, the longitudinal areas run transversely, preferably at an angle of greater than or equal to 85° to less than or equal to 90° to the defined transverse direction. The longitudinal areas preferably run perpendicularly to the transverse areas or parallel to the plant row center line and thus form a kind of grid structure in the image information. However, the longitudinal areas can also run vertically in the image information or the image, independently of the defined transverse direction or the plant row center line. The longitudinal regions preferably have an essentially equal length in the defined transverse direction. In this case, the length of the longitudinal regions is greater than 1. The length of the longitudinal regions is preferably in a range from greater than or equal to 5 pixels to less than or equal to 50 pixels.

The alignment of the longitudinal areas and the transverse areas relative to the identified plant row center line offers the advantage, in particular compared to the known methods with bounding boxes, that the weeds, regardless of the arrangement of the plant row center line in the image information or the image, which, for example, also at an angle of 45°, can be recognized very well within the corresponding plant rows.

The defined plant portion is a longitudinally defined portion of an identified plant of the plant row. A plant section can include the entire identified plant or the entire plant body in the longitudinal direction. However, a plant portion may include only a longitudinal portion of an identified plant. Thus, a plant can be divided into two or more plant sections. The defined plant section is also bounded on one side in the transverse direction by the plant center line. Accordingly, a defined plant section, which is to the left of the plant center line, is delimited by it on the right, while a defined plant section, which is to the right of the plant center line, is delimited by it on the left.

In this case, in the step of determining the expansion parameters, the plant sections are defined or formed in the longitudinal direction in that an adjacent transverse area in the respective crop area is empty, or that a change in the partial width value from a transverse area to the partial width value of an adjacent transverse area is greater than or equal to a defined change threshold value .

The procedure here is preferably such that as soon as a plant or a plant part or plant mass is identified in a transverse area, the counting begins and only ends when—seen in the longitudinal direction—a subsequent transverse area is empty or the partial width values of two adjacent transverse areas deviate from one another in this way that a defined change threshold is reached or exceeded. The change threshold value can be specified here. The term “empty” means that no plant or part of a plant was identified in the transverse area in the respective crop area. This determines the length of the plant sections in the unit “transverse areas”.

A section value represents the number of longitudinal areas over which a plant section extends in a single transverse area, starting from the defined plant row center line. A section value can thus be assigned to each transverse area.
The extent index is preferably a quotient of the width value, ie the sum of all partial width values of a plant section, to the length value of the respective plant section. In this way, each plant section can be assigned an expansion index.

In the step of determining the expansion parameters, preferably only plants are taken into account, ie filtered out and evaluated, which are at least partially arranged in the respective defined crop area, ie are assigned to the crop area.

The method also includes a step of comparing the determined extent index of the respective defined plant section with a frequency distribution, i.e. a history of extent indexes of defined plant sections of previously identified plants of the defined crop area and/or defined crop areas of previously recorded image information of the agricultural area by means of the computing unit.

The defined crop plant areas of the previously recorded image information preferably include the same at least one identified row of plants or previously recorded sections of the same row of plants. Accordingly, the previously captured image information is preferably image information captured immediately beforehand by the image capturing unit, in particular during the same movement or crossing of the image capturing unit over the agricultural area.

The frequency distribution is preferably a distribution of the absolute frequency of the extent metrics. However, it is possible that the frequency distribution is a relative frequency, e.g. based on the sum of 300 plant objects. Understandably, 2 or more extension key figures in the frequency distribution can also be combined into key figure classes depending on a specific parameter, without departing from the scope of the present application.

The method also includes a step of identifying the plant sections with an extent index whose frequency is less than or equal to a defined frequency threshold value of the frequency distribution in order to identify the plants belonging to the plant sections as weeds in an identified plant row.

The defined frequency threshold of the frequency distribution can be specified. The defined frequency threshold value can depend, for example, on the species and/or the genus and/or the growth stage and/or the distribution of the area/size of the useful plants. However, the defined frequency threshold can also be adjusted depending on a specific parameter.

It is advantageous if the frequency distribution is dynamically adapted with an increasing number of previously identified plants or a defined number of plants identified immediately previously during the method, in particular during a movement of the optical detection unit over the agricultural area. This means, in other words, that the frequency distribution changes dynamically while crossing the field, since (new) length values and (new) width values and thus (new) dimensions of newly identified plants are always included. Furthermore, (old) longitude values and (old) latitude values and thus (old) extent parameters of old identified plants are preferably dropped, e.g. if only a defined constant number of currently identified plants is taken into account, so that the frequency distribution adapts to the latest conditions on the agricultural area . This measure makes the method very flexible, since the histogram is constantly renewed when crossing the agricultural area and the oldest data is removed from the histogram again, so that the algorithm can constantly relearn the current field situation and extent of the plants and is also dynamic enough that he can adapt his decision regarding the control of the agricultural work tool even when plant growth changes across the agricultural area.

In the step of identifying the plant sections, only plants are preferably taken into account, i.e. filtered out and evaluated, which are at least partially arranged in the respective defined crop plant area, i.e. are assigned to the crop plant area.

It is advantageous if, in the step of identifying, only plants with an extent index that is greater than or equal to a defined index threshold value are identified as weeds in an identified row of plants. The defined index threshold value can advantageously be specified or determined as a function of the frequency distribution of expansion indexes. It is advantageous here if, when determining the key figure threshold value, the largest extent key figure is determined with a frequency above the frequency threshold value, and the key figure threshold value is defined as the next larger extent key figure or the larger or smaller extent key figure by a defined value. The defined value by which the index threshold value should be greater may depend on the species and/or the genus and/or the growth stage and/or the spread of the area/size of the crop plants. With this measure, large weeds in particular, ie, weeds that have a greater width-to-length ratio than the plants or crops in the plant row can be effectively identified.

A minimum value can also be defined for the key figure threshold value, above which a plant may be identified/classified as a weed or reclassified from the crop plant class into a weed class. The minimum value can depend, for example, on the species and/or the genus and/or the growth stage and/or the distribution of the area/size of the useful plants. This is a kind of security so that the dynamic and self-learning algorithm only reclassifies larger weeds. According to this strategy, smaller weeds with a smaller key figure threshold value should remain unchanged, e.g. if the crop plants are smaller and only larger weeds should be identified.

It is also advantageous if the method includes a step of classifying the identified weeds in the identified row of plants into a plant class, in particular a weed class, by means of the computing unit. It is particularly advantageous here if a step of classifying the, i.e. all identified plants of the defined crop plant area into a first plant class, in particular a crop plant class, is initially provided, and then a step of reclassifying the identified weeds in the identified plant row of the defined crop plant area into a second plant class different from the first plant class, in particular a weed class, is provided by means of the computing unit. This means, in other words, that initially, by definition, all plants in the crop area are viewed as objects of the same class or are assigned to the same class, e.g. the class "crop plants", and then misclassified weeds within the plant rows are identified and correctly reclassified.

Furthermore, it is advantageous if the method includes a step of controlling an agricultural working machine, in particular an agricultural working tool of an agricultural working machine, in particular a spray device of a field sprayer, depending on the identified weeds in the identified row of plants. As a result, the method makes it possible to combat or regulate weeds in the plant rows in a targeted manner.

The arithmetic unit is or the arithmetic 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 computing unit has corresponding image processing software. The arithmetic unit can be, for example, a signal processor, a microcontroller or the like, with the memory unit being able to be a flash memory, an EPROM or a magnetic memory 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 processing unit or a control unit.

The processing unit can be arranged completely or partially on the agricultural working machine or integrated into it. However, the computing unit can also be completely or partially external, for example integrated in a cloud. The arithmetic unit can thus also be divided among different units, for example mobile and stationary units.

A computer program product or computer program with program code, which can be stored on a machine-readable carrier or storage medium such as a semiconductor memory, a hard disk memory or an optical memory and for carrying out, implementing and/or controlling the steps of the method according to one of the embodiments described above, is also advantageous used, especially when the program product or program is run on a computer or device.

• 1 eine Bildinformation bzw. ein Bild eines erfassten Feldabschnitts;
• 2 eine Bildinformation mit identifizierten Pflanzen, welche in Querbereiche und Längsbereiche unterteilt ist;
• 3 eine Historie von Pflanzenabschnitten aus zuvor erfassten Bildinformationen mit ihren Ausdehnungskennzahlen;
• 4 eine Bildinformation mit einem aufgrund der Überschreitung des Änderungsschwellenwerts definierten Pflanzenabschnitt; und
• 5 ein Ablaufdiagramm eines Verfahrens gemäß einem Ausführungsbeispiel.

Embodiments of the invention are shown in the drawings and explained in more detail in the following description. It shows:

• 1 an item of image information or an image of a detected field section;
• 2 an image information with identified plants, which is divided into transverse areas and longitudinal areas;
• 3 a history of plant sections from previously acquired image information with their extent indices;
• 4 a piece of image information with a plant section defined as a result of the change threshold value being exceeded; and
• 5 a flow chart of a method according to an embodiment.

1 shows image information or an image 10 of a field section 12 of an agricultural area with plants 14 captured by means of an optical detection unit or camera (not shown). The plants 14 include crop plants 16 and weeds 18.

How out 1 As can be seen, after the plants 14 had been detected (using the red color component and/or infrared component and plant distance information) in the field section 12, rows of plants 20 were identified by means of a computing unit (not shown). The plant rows 20 were identified by fitting defined straight plant row center lines 22 in image trajectories with the highest NDVI value (difference red to NIR) of the plants 14 .

How further from 1 As can be seen, a cultivated plant area 24, which includes the identified plant rows 20, was defined by the computing unit. The cultivated plant area 24 is tubular and has a defined, constant width around the respective defined plant row center line 22 . The cultivated plant area 24 does not include the entire plant body of the plants 14 of the plant rows 20.

2 now illustrates how in the image information 10 in each case an expansion index for defined plant sections 26 is determined. The image information 10 or the image 10 is divided into transverse areas 28 and longitudinal areas 30 . The transverse areas 28 are rectangles of a defined length 32 which are spanned over the entire length of the image 10 in a longitudinal direction 34 which runs along the center line 22 of the plant rows. The transverse regions 28 run perpendicular to the longitudinal direction 34 or plant row center line 22. The longitudinal regions 30 are also rectangles of a defined width 36, which are spanned across the entire width of the image 10 in a transverse direction 38, which runs perpendicular to the plant row center line 22. The longitudinal areas 30 run perpendicularly to the transverse direction 38 or parallel to the plant row center line 22.

First, for each row row 20 or each crop area 24, plants 14, 16, 18 are filtered out, which are at least partially arranged in the respective defined crop area 24. Consequently, only these filtered out plants 16, 18 are subsequently taken into account and evaluated.

An extension index for defined plant sections 26 of these plants 16, 18 is then determined by the computing unit, with the extension index representing a ratio or a quotient of a width value and a length value of the respective defined plant section 26. In this case, the length value represents the number of transverse regions 28 over which the respective defined plant section 26 extends in the defined longitudinal direction 34 . The width value is in turn a sum of part-width values, with the part-width values representing the number of longitudinal regions 30 over which the respective defined plant section 26 extends in the respective transverse regions 28, starting from the plant row center line 22 in the defined transverse direction 38.

Here, a plant section 26 begins by definition as soon as plant matter is identified in a transverse area 28 or a change in a section value from one transverse area 28 to the section value of an adjacent transverse area 28 is greater than or equal to a defined change threshold value. Similarly, the crop section 26 ends when a transverse region 28 is empty, i.e. no more vegetation is identified or a change in a section value from a transverse region 28 to the section value of an adjacent transverse region 28 is greater than or equal to a defined change threshold value.

This gives plant sections 26 with a corresponding length value, namely the number of transverse regions 28, and a corresponding width value, namely the sum of the number of longitudinal regions 30. In the exemplary embodiment shown, a plant section 26 is attached to the plant 16 to the left and right of the plant row center line 22 Are defined. The left plant section 26 of the plant 16 has the length value 2 and the width value 2+1=3. This corresponds to an expansion index of 3/2=1.5. The right plant section 26 of the plant 16 has a length value of 1 and a width value of 1. This corresponds to an expansion index of 1/1=1. Analogous to this, the left plant section 26 of the plant 18 has the length value 2 and the width value 3+3=6. This corresponds to an expansion index of 6/2=3. The right plant section 26 of the plant 18 has the length value 5 and the width value 4+7+8+4+4=27. This corresponds to an expansion index of 27/5=5.4. In this case, depending on the change threshold value, right plant section 26 of plant 18 could also have been divided into three plant sections, for example, if this were to be 3, since a division or separation when changing the section width values from 4 to 7 and from 8 to 4.

3 shows another example of image information 10 with a defined plant section 26, which was defined or formed on both sides in the longitudinal direction 34 on the basis of a corresponding change threshold value.

The expansion characteristics or data are continuously stored during a crossing over the agricultural area. A history is built up in which the last, for example 100, plants 14, 16, 18 or plant sections 26 from the previously recorded image information 10 are stored. In 4 1 shows the frequency distribution or history of the determined expansion parameters using a bar chart, which is always dynamically adapted based on new image information 10 . The expansion parameters are plotted on the abscissa and the number of plant sections 26 with the respective expansion index are plotted on the ordinate. Using the data collected from the past, it is now possible to estimate which plant sections 26 or plants 14, 16, 18 occurred more frequently with which expansion parameters, i.e. are cultivated plants 16, and which plant sections 26 or plants 14, 16, 18 occurred less frequently, i.e. large weeds 18 are.

In order to identify the weeds 18 in the identified row of plants 20, i.e. in the cultivated plant area 24 comprising the row of plants identified 20, a frequency threshold value 40 was defined in the frequency distribution, which in the example shown is 26 for 3 plant sections.

In a first exemplary embodiment, all plant sections 26 or plants 14 of the defined cultivated plant area 24 whose frequency is less than or equal to the defined frequency threshold value 40, i.e. occur 3 times or less, are identified or classified as weeds 18. Thus, in this exemplary embodiment, plant sections 26 or plants 14 with an extension index of 1.75, 2.5 and 2.75 are identified as weeds 18 . All remaining plant sections 26 or plants 14 with a frequency of more than 3 are identified or classified as crop plants 16 .

In a second exemplary embodiment, an index threshold value 42 is also defined, with only plant sections 26 or plants 14 having an extent index greater than or equal to this index threshold value 42 being identified as weeds 18 . Here, the metric threshold was determined by first determining the largest extent metric with a frequency above the frequency threshold, which is 2.25 in the present example. The next largest expansion index, i.e. 2.5, was then selected as the index threshold value.

Thus, compared to the first embodiment, in this embodiment only plant sections 26 or plants 14 with an expansion index of 2.5 and 2.75 are identified or classified as weeds 18, while the plant sections 26 or plants 14 with an expansion index of 1, 75 and all others are identified or classified as cultivated plants 16 . As a result, large weeds 18 can be effectively identified and, if necessary, reclassified.

In a third exemplary embodiment, in comparison to the second exemplary embodiment, a characteristic value threshold value 44 can be defined, which is greater by a defined value than the ascertained greatest characteristic value for expansion. As a result, only wider plants 14 with an expansion index of 2.75 are identified or classified as weeds 18 .

5 shows a flow chart of an embodiment of the approach presented here as a method 100 for identifying weeds 18 in a row of plants 20 of an agricultural area. Method 100 includes a step of receiving 102 image information 10 from a field section 12 of an agricultural area with plants 14, 16, 18 that is detected by means of an optical detection unit. Method 100 also includes a step of identifying 104 at least one row of plants 20 in the image information 10 using identified plants 14, 16, 18 by means of a computing unit. The method 100 also includes a step of defining 106 a crop plant region 24 comprising the at least one identified row of plants 20 using the at least one identified row of plants 20 in the image information 10 by means of the computing unit. The method 100 further comprises a step of determining 110 an extent index for defined plant sections 26 of the identified plants 14, 16, 18 of the defined cultivated plant area 24 in the image information 10 by means of the computing unit, the extent index being a ratio of a latitude value and a longitude value of the respective defined plant section 26, the image information 10 being divided into transverse regions 28 and longitudinal regions 30, the length value representing the number of transverse regions 28 over which the respective defined plant section 26 extends in a defined longitudinal direction 34, and the width value representing one Sum of partial width values, the partial width values representing the number of longitudinal regions 30 over which the respective defined plant section 26 extends in the respective transverse regions 28, starting from a defined plant row center line 22 of the respective plant row 20 in a defined transverse direction 38. The method 100 also includes a step of comparing 112 the determined extent index of the respective defined plant section 26 with a frequency distribution of extent indexes of defined plant sections 26 of previously identified plants 14, 16, 18 of the defined crop area 24 and/or defined crop areas (24) of previously recorded image information 10 of the agricultural area by means of the unit of account. The method 100 also includes a step of identifying 114 the plant sections 26 with an extent index whose frequency is less than or equal to a defined frequency threshold value 40 of the frequency distribution, in order to identify the plants 14 associated with the plant sections 26 as weeds 18 in an identified plant row 20.

Furthermore, the method 100 comprises an optional step of classifying 108 the identified plants 14, 16, 18 of the defined crop region 24 into a first plant class, in particular a crop plant class, and a further optional step of reclassifying 116 the identified weeds 18 in the identified plant row 20 of the defined cultivated plant area 24 into a second plant class different from the first plant class, in particular a weed class, by means of the computing unit.

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 [0002]

Claims (18)

Method (100) for identifying weeds (18) in a plant row (20) of an agricultural area, with the steps: - receiving (102) image information (10) from a field section (12) of an agricultural area with plants (14, 16, 18) detected by means of an optical detection unit; - Identification (104) of at least one row of plants (20) in the image information (10) using identified plants (14, 16, 18) by means of a computing unit; - defining (106) a crop plant region (24) comprising the at least one identified row of plants (20) using the at least one identified row of plants (20) in the image information (10) by means of the computing unit; - Determination (110) in each case of an extent index for defined plant sections (26) of the identified plants (14, 16, 18) of the defined cultivated plant area (24) in the image information (10) by means of the computing unit, the extent index being a ratio of a width value and a length value of the respective defined plant section (26), the image information (10) being divided into transverse regions (28) and longitudinal regions (30), the length value representing the number of transverse regions (28) over which the respective defined plant section (26) extends extends in a defined longitudinal direction (34), and the width value is a sum of partial width values, the partial width values being the number of longitudinal regions (30) over which the respective defined plant section (26) extends in the respective transverse regions (28) starting from a defined plant row center line (22) of the respective plant row (20) extending in a defined transverse direction (38); - Comparing (112) the determined extent index of the respective defined plant section (26) with a frequency distribution of extent indexes of defined plant sections (26) of previously identified plants (14, 16, 18) of the defined crop plant area (24) and/or defined crop plant areas (24) previously captured image information (10) of the agricultural area by means of the computing unit; and - Identifying (114) the plant sections (26) with an extent index whose frequency is less than or equal to a defined frequency threshold value (40) of the frequency distribution, around the plants (14) belonging to the plant sections (26) as weeds (18) in an identified plant row (20) to identify. Method (100) according to claim 1 , characterized in that in the step of determining (110) the expansion parameters, the plant sections (26) are defined or formed in the longitudinal direction (34) in that an adjacent transverse area (28) in the respective crop area (24) is empty, or that a change in the partial width value from a transverse area (28) to the partial width value of an adjacent transverse area (28) is greater than or equal to a defined change threshold value. Method (100) according to claim 1 or 2 , characterized in that the expansion index is a quotient of the width value to the length value of the respective plant section (26). Method (100) according to one of the preceding claims, characterized in that in the step of identifying (114) also only plant sections (26) with an expansion index which is greater than or equal to a defined index threshold value (42) as weeds (18) in a identified plant series (20) are identified. Method (100) according to claim 4 , characterized in that the defined index threshold value (42) is predetermined or determined as a function of the frequency distribution of expansion indexes. Method (100) according to claim 5 , characterized in that when determining the index threshold value (42), the largest extent index is determined with a frequency above the frequency threshold value (40), and the index threshold value (42) is defined as the next larger extent index or the extent index larger or smaller by a defined value . Method (100) according to one of the preceding claims, characterized in that only plants (14, 16, 18) are taken into account in the step of determining (110) the extent parameters and/or in the step of identifying (114) the plant sections (26), which are at least partially arranged in the respective defined crop area (24). Method (100) according to one of the preceding claims, characterized in that the defined longitudinal direction (34) along and the defined transverse direction (38) transversely, in particular at an angle of greater than or equal to 85° to less than or equal to 90° to the plant row center line ( 22) of the identified plant row (20), in particular wherein the transverse areas (28) have a substantially equal length and/or the longitudinal areas (30) have a substantially equal width. Method (100) according to one of the preceding claims, characterized in that the defined crop plant areas (24) of the previously recorded image information (10) comprise the same at least one identified row of plants (20). Method (100) according to one of the preceding claims, characterized in that the frequency distribution with an increasing number of previously identified plants (14) or a defined number of immediately previously identified plants (14) during the method, in particular during a movement of the optical detection unit over the agricultural surface is dynamically adjusted. Method (100) according to one of the preceding claims, characterized by a step of classifying (116) the identified weeds (18) in the identified plant row (20) into a plant class, in particular a weed class, by means of the computing unit. Method (100) according to claim 11 , characterized by a step of classifying (108) the identified plants (14, 16, 18) of the defined crop plant area (24) into a first plant class, in particular a crop plant class, and a step of reclassifying (116) the identified weeds (18) in the identified row of plants (20) of the defined cultivated plant area (24) into a second plant class that differs from the first plant class, in particular a weed class, by means of the computing unit. Method (100) according to one of the preceding claims, characterized by a step of controlling (118) an agricultural working machine, in particular an agricultural working tool of an agricultural working machine, in particular a spray device of a field sprayer, depending on the identified weeds (18) in the identified row of plants (20). Arithmetic unit which is set up to carry out and/or to control the steps of a method (100) according to one of the preceding claims. Plant identification unit with an optical detection unit for detecting a field section (12) of an agricultural area with plants (14, 16, 18) in order to obtain image information (10) from the detected field section (12), and a computing unit Claim 14 . Agricultural working machine, in particular field sprayer, with an agricultural working tool, in particular a spray device, and a plant identification unit claim 15 , wherein the working tool, in particular the spraying device, is controlled depending on the identified weeds (18) in the identified row of plants (20). Computer program that is set up to carry out the steps of a method (100) according to one of Claims 1 until 13 perform and/or control when the computer program is run on a computer. Machine-readable storage medium on which the computer program Claim 17 is saved.

Images