WO2021249560A1 - 作物缺失的检测方法及检测装置 - Google Patents

作物缺失的检测方法及检测装置 Download PDF

Info

Publication number
WO2021249560A1
WO2021249560A1 PCT/CN2021/099863 CN2021099863W WO2021249560A1 WO 2021249560 A1 WO2021249560 A1 WO 2021249560A1 CN 2021099863 W CN2021099863 W CN 2021099863W WO 2021249560 A1 WO2021249560 A1 WO 2021249560A1
Authority
WO
WIPO (PCT)
Prior art keywords
crop
area
planting row
vegetation
image
Prior art date
Application number
PCT/CN2021/099863
Other languages
English (en)
French (fr)
Inventor
黄敬易
Original Assignee
广州极飞科技股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from CN202010537090.3A external-priority patent/CN113807128B/zh
Priority claimed from CN202010538238.5A external-priority patent/CN113807138A/zh
Application filed by 广州极飞科技股份有限公司 filed Critical 广州极飞科技股份有限公司
Publication of WO2021249560A1 publication Critical patent/WO2021249560A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition

Definitions

  • This application relates to the field of crop identification, and in particular to a method and device for detecting crop absence.
  • the embodiments of the present application provide a method and device for detecting missing crops, so as to at least solve the technical problems that manual detection of missing crops requires a lot of manpower, is time-consuming and laborious, and is prone to human error.
  • a method for detecting crop absence including: obtaining a regional image of a target area; extracting a vegetation area from the regional image, and generating a first binarized image corresponding to the vegetation area; Determine the crop connected domain in the vegetation connected domain of the first binarized image, and generate a second binarized image based on the crop connected domain; for each planting row area in the second binarized image, determine the planting row area
  • the crop distance between any two adjacent crop connected areas, and/or the critical distance between the end crop connected areas in the planting row area and the boundary of the planting row area, where the end crop connected areas are in the planting row area In is only adjacent to one crop connected domain; according to the crop spacing or critical spacing, it is determined whether the planting row area is in a crop missing state.
  • determining whether the planting row area is in a crop missing state according to the crop spacing includes: performing bidirectional detection of the planting row area according to a preset threshold, where the bidirectional detection is based on one direction for each crop connected domain in the planting row area After comparing the distance between each crop with a preset threshold, then compare the distance between the connected domains of each crop with the preset threshold in the opposite direction of the direction; when the distance is greater than the preset threshold, it is determined that the planting row area is in a crop missing state. Including: For any two adjacent crop connected domains, when the detection result in any direction in the bidirectional detection result indicates that the distance between any two adjacent crop connected domains is greater than a preset threshold, it is determined that the planting row area is in the crop Missing status.
  • determining whether the planting row area is in a crop missing state according to the critical distance includes: obtaining a crop boundary that matches the second binarized image; obtaining an end crop boundary that matches the connected domain of the end crop from the crop boundary; Along the main planting row direction of the planting row area, calculate the critical distance between the center of mass of the connected area of the end crop and the boundary of the end crop; when the critical distance is greater than a preset threshold, it is determined that the planting row area is in a crop missing state.
  • the process of determining each planting row area in the second binary image includes: determining the main direction of the crop planting row in the target area; generating the first binary value based on the length of each planting row in the target area in the main direction
  • the circumscribed rectangle of the chemical image and use the width of the rectangle as the ordinate and the length as the abscissa to establish a coordinate system; determine the number of non-vegetation pixels in the main direction, and generate it in the coordinate system based on the number of non-vegetation pixels
  • a curve used to indicate the distribution of the number of non-vegetation pixels according to the peak apex of the curve and the line in the line set corresponding to the main direction, the second binarized image is cut to obtain the planting row area, where the line in the line set It is used to indicate the set of location points where non-vegetation pixels are located between adjacent planting rows.
  • determining the main direction of the crop planting rows in the target area includes: performing Hough transform on the second binarized image, and outputting the extremes of the second binarized image in the Hough space through a Hough space accumulator.
  • the coordinate system parameter pair list based on the angle parameter in the polar coordinate system parameter pair list, determines the main direction of the crop planting row, where the polar coordinate system parameter pair list includes at least one polar coordinate system parameter pair, and each polar coordinate system parameter The centering includes angle and radius.
  • output the polar coordinate system parameter pair list of the second binarized image in the Hough space through the Hough space accumulator and determine the main direction of the crop planting row based on the angle parameter in the polar coordinate system parameter pair list, including : When the number of polar coordinate parameter pairs in the polar coordinate system parameter pair list is greater than the predetermined value, the angle that appears most frequently in the polar coordinate parameter pair list is used as the main direction of the crop planting row; or, in the polar coordinate system parameter pair list When the number of polar coordinate parameter pairs in is less than the predetermined value, the polar coordinate angle at the top of the list of polar coordinate system parameter pairs is taken as the main direction of the crop planting row.
  • the method before cutting the second binarized image to obtain the planting row area according to the peak apex of the curve and the straight line in the line set corresponding to the main direction, the method further includes: smoothing the curve to obtain a smoothed target curve ; When non-vegetation pixels are included in the smoothed curve, offset the set of straight lines so that the target curve does not include non-vegetation pixels.
  • cutting the second binarized image to obtain the planting row area according to the line in the set of straight lines corresponding to the peak apex of the curve and the main direction includes: obtaining the abscissa of each peak apex in the curve, and following the main direction of the planting line Direction generates straight lines; in the second binary graph, the area between two adjacent straight lines is used as the planting row area.
  • the abscissa of each peak apex in the curve further includes: sequentially obtaining the abscissa of a target peak apex as the current processing abscissa, and obtaining the neighborhood abscissa set associated with the current processing abscissa; if If the abscissa of the target neighborhood in the neighborhood abscissa set meets the critical point condition, replace the abscissa of the target peak apex with the abscissa of the target neighborhood; return to the operation of obtaining the abscissa of a target peak apex as the current processing abscissa in turn , Until the processing of the abscissa of all peak vertices is completed.
  • the method further includes: extracting samples from the second binarized image Data, where the sample data includes: crop connected domains; label the crop connected domains in the sample data to obtain the crop label; train the crop recognition model based on the crop connected domains and the labels assigned to the crop connected domains to obtain training After the crop recognition model.
  • extracting the vegetation area from the regional image and generating the first binary image corresponding to the vegetation area includes: determining the greenness index of each pixel in the regional image; for each pixel, comparing the greenness index and The size of the greenness threshold; determine whether the pixel is a pixel in the vegetation area according to the comparison result; count the pixels in the vegetation area, and determine the first binary image based on the statistical result.
  • determining the crop connected domains from the vegetation connected domains of the first binarized image, and generating a second binarized image based on the crop connected domains includes: obtaining information about each vegetation connected domain in the first binarized image Area; for each vegetation connected domain, determine whether the area of the vegetation connected domain belongs to the preset value range, if it belongs to the preset value range, determine the vegetation connected domain as a crop connected domain; generate a binary value based on the determined crop connected domain ⁇ image.
  • the preset threshold can be obtained by any of the following methods: The first method: Count the set of distances between the centroids of any two adjacent crop connected areas in the target area; calculate the average of all distances in the distance set , And determine a preset threshold based on the average value, where the distance is the distance between the centroids of adjacent crop connected domains; the second way: receiving the preset threshold input by the user.
  • determining the preset threshold value based on the average value includes: determining the average value as the preset threshold value; or, calculating the radius of each crop connected area in the target area to obtain the average radius of the crop connected area, based on the average radius and the average value Determine the preset threshold.
  • the method further includes: determining a position in the planting row area in the crop missing state; and generating a marker for indicating the crop missing state at the position in the crop missing state.
  • generating a mark for indicating the missing state of the crop at the position in the missing state of the crop includes: generating a circular mark with the position in the missing state as the center and the average radius of the connected domain as the radius, and combining the The circular mark is used as a mark to indicate the missing state of crops; or a rectangular mark is generated with the position in the missing state as the center and twice the average radius of the connected domain as the side length, and the rectangular mark is used to indicate the missing crop Status mark.
  • acquiring an area image of the target area includes: receiving an area image of the target area taken by a drone.
  • a method for detecting crop absence which includes: acquiring a region image of a target region; inputting the region image to a crop recognition model for analysis, and obtaining each crop in the target region Connected domains, and generate binarized images based on the connected domains of crops.
  • the crop recognition model is trained based on multiple sets of data. Each set of data in the multiple sets of data includes sample images and labels used to mark the sample images.
  • Connected domain label determine the crop spacing between any two adjacent crop connected domains in each planting row area in the binarized image, and/or the boundary between the end crop connected domains in the planting row area and the planting row area Among them, the end crop connected domain is adjacent to only one crop connected domain in the planting row area; according to the crop distance or the critical distance, it is determined whether the planting row area is in a crop missing state.
  • determining whether the planting row area is in a crop missing state according to the crop spacing includes: performing bidirectional detection of the planting row area according to a preset threshold, where the bidirectional detection is based on one direction for each crop connected domain in the planting row area After comparing the distance between each crop with a preset threshold, the distance between each crop connected domain is compared with the preset threshold in the opposite direction of the direction; for any two adjacent crop connected domains, in the two-way detection result When the detection result in any one direction indicates that the crop distance between any two adjacent crop connected domains is greater than a preset threshold, it is determined that the planting row area is in a crop missing state.
  • the process of determining each planting row area in the binarized image includes: determining the main direction of the crop planting row in the target area; generating the circumscribed image of the binarized image based on the length of each planting row in the target area in the main direction Rectangular; the width of the rectangle is the ordinate and the length is the abscissa to establish a coordinate system; determine the number of non-vegetation pixels in the main direction, and based on the number of non-vegetation pixels, generate in the coordinate system to indicate non-vegetation
  • the curve of the number distribution of vegetation pixels according to the peak apex of the curve and the straight line in the straight line set corresponding to the main direction, the second binarized image is cut to obtain the planting row area, where the straight line in the straight line set is used to indicate the phase A collection of location points where non-vegetation pixels are located between adjacent planting rows.
  • a method for calculating the number of missing crops including: based on the crop missing detection method, processing the regional image of the target area to determine whether the planting row area contained in the target area is in the crop Missing status; when it is determined that the planting row area is in a crop missing state, a crop missing mark is generated; based on the crop missing mark, the number of missing crops is calculated.
  • a method for planning an operation route which includes: determining an area lacking seedlings based on a detection method for crop missing; generating an operation route according to the area lacking seedlings, wherein the operation route passes through the area lacking seedlings.
  • a method for replanting including: determining the lack of seedling area based on the detection method of crop lack; generating the operation route according to the lack of seedling area; sending the operation route to the operation equipment, wherein, the operation route Used to instruct operating equipment to replant crops in areas lacking seedlings.
  • the process of determining the area lacking seedlings includes: processing the area image of the target area based on the crop missing detection method to determine whether the planting row area contained in the target area is in a crop missing state; determining that the planting row area is in a crop missing state In the state, the crop missing marker is generated; according to the image position of the crop missing marker in the regional image binary map and the geographic location information matching the regional image binary map, the geographic location information corresponding to each crop missing marker is determined to obtain the defect.
  • Miao area processing the area image of the target area based on the crop missing detection method to determine whether the planting row area contained in the target area is in a crop missing state; determining that the planting row area is in a crop missing state In the state, the crop missing marker is generated; according to the image position of the crop missing marker in the regional image binary map and the geographic location information matching the regional image binary map, the geographic location information corresponding to each crop missing marker is determined to obtain the defect.
  • generating a replanting operation route according to the seedling shortage area includes: generating a replanting operation route matching the planting area according to geographic location information corresponding to the crop lacking seedling mark.
  • a working method which is applied to working equipment, and the working equipment includes at least one of the following: spraying equipment, spreading equipment, and harvesting equipment; the method includes: determining the current working equipment Whether the position is located at the seedling shortage position of the target area, if the operating equipment is at the seedling shortage position, stop working at the seedling shortage position, where the seedling shortage position is determined based on the crop lacking detection method.
  • a yield measurement method including: determining a seedling shortage area in a target area based on a detection method of crop missing; determining the area of a non-seeding area in the target area according to the seedling shortage area ; Determine the total yield of the target area based on the yield per unit area of the non-deficient area and the total area of the non-deficiency area.
  • a device for detecting crop absence in a target area including: an acquisition module configured to acquire an area image of the target area; a first generation module configured to extract an area image from the area image Vegetation area, and generate the first binary image corresponding to the vegetation area; the second generation module is set to determine the crop connected domain from the vegetation connected domain of the first binary image, and generate the second second binary image based on the crop connected domain Valued image; the first determination module is set to determine the crop spacing between any two adjacent crop connected domains in the planting row area for each planting row area in the second binarized image, and/or planting The critical distance between the end crop connected domain and the border of the planting row area in the row area, where the end crop connected area is adjacent to only one crop connected area in the planting row area; the second determining module is set to be based on the crop The spacing or critical spacing determines whether the planting row area is in a crop missing state.
  • an unmanned aerial vehicle including: an image acquisition device configured to acquire an area image of a target area; a processor configured to extract a vegetation area from the area image and generate vegetation For the first binarized image corresponding to the region, determine the crop connected domain from the vegetation connected domain of the first binarized image, and generate a second binarized image based on the crop connected domain; for the second binarized image For each planting row area, determine the crop spacing between any two adjacent crop connected areas in the planting row area, and/or the critical distance between the end crop connected areas in the planting row area and the boundary of the planting row area, Among them, the end crop connected domain is adjacent to only one crop connected domain in the planting row area; whether the planting row area is in a crop missing state is determined according to the crop spacing or the critical spacing.
  • a device for calculating the number of missing crops including: an acquisition module configured to acquire an area image of a target area; a recognition module configured to extract a vegetation area from the area image, and Generate the first binarized image corresponding to the vegetation area; determine the crop connected domain from the vegetation connected domain of the first binarized image, and generate a second binarized image based on the crop connected domain; for the second binarized image For each planting row area in the planting row area, determine the crop spacing between any two adjacent crop connected areas in the planting row area, and/or the boundary between the end crop connected areas in the planting row area and the planting row area Distance; where the end crop connected domain is adjacent to only one crop connected domain in the planting row area; the calculation module is set to determine whether the planting row area is in a crop missing state according to the crop distance or critical distance.
  • a non-volatile storage medium includes a stored program, wherein, when the program is running, the device where the non-volatile storage medium is located is controlled to execute Any method of detecting missing crops, or calculating method of missing crops, or planning method of operation path, or method of replanting, or method of operation, or calculation of output method.
  • a processor wherein the processor is used to run a program, wherein, when the program is running, any method for detecting missing crops or calculating the number of missing crops is executed.
  • the method of identifying the region image based on the deep learning algorithm is adopted to obtain the region image of the target region; the vegetation region is extracted from the region image, and the first binary image corresponding to the vegetation region is generated; Determine the crop connected domain in the vegetation connected domain of the binarized image, and generate a second binarized image based on the crop connected domain; for each planting row area in the second binarized image, determine any two of the planting row areas The crop distance between adjacent crop connected domains, and/or the critical distance between the end crop connected domains in the planting row area and the boundary of the planting row area, wherein the end crop connected domains are only in the planting row area.
  • Adjacent to a crop connection domain determine whether the planting row area is in a crop missing state according to the crop spacing or critical spacing, achieving the purpose of identifying missing crops, thus achieving the technical effect of timely, efficient and accurate identification of missing crops, and then solving
  • the manual detection of missing crops requires a lot of human participation, time-consuming and labor-intensive, and is prone to technical problems such as human error.
  • Fig. 1a is a schematic flowchart of a method for detecting crop absence according to an embodiment of the present application
  • Figure 1b is a schematic diagram of a planting row area according to an embodiment of the present application.
  • FIG. 2 is a schematic diagram of an optional coordinate system established with the width of a circumscribed rectangle as the ordinate according to an embodiment of the present application;
  • Fig. 3 is a schematic diagram of an optional curve segmentation frame according to an embodiment of the present application.
  • Fig. 4 is a schematic diagram of an optional smooth curve frame according to an embodiment of the present application.
  • Fig. 5 is a schematic structural diagram of a device for detecting crop absence according to an embodiment of the present application.
  • Fig. 6 is a schematic structural diagram of an optional drone according to an embodiment of the present application.
  • Fig. 7 is a schematic flowchart of a method for calculating the number of missing crops according to an embodiment of the present application.
  • Fig. 8 is a schematic structural diagram of a device for calculating the number of missing crops according to an embodiment of the present application.
  • FIG. 9 is a schematic flowchart of another method for detecting crop absence according to an embodiment of the present application.
  • Fig. 10 is a schematic diagram of an optional detection principle of crop missing according to an embodiment of the present application.
  • an embodiment of a method for detecting crop absence is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and Although the logical sequence is shown in the flowchart, in some cases, the steps shown or described can be performed in a different order than here.
  • Fig. 1a is a method for detecting crop absence according to an embodiment of the present application. As shown in Fig. 1a, the method includes the following steps:
  • Step S102 acquiring an area image of the target area
  • Step S104 extracting the vegetation area from the area image, and generating a first binary image corresponding to the vegetation area;
  • the implementation of extracting the vegetation area includes, but is not limited to: generating a vegetation binary image, that is, the first binary image, according to the vegetation points included in the area image. specifically:
  • Vegetation points indicate that the pixels in the image represent vegetation, and whether the pixels in the image are vegetation points can be determined by means of color space conversion, color index, and vegetation index.
  • the Excess Green Index (EXG) in the vegetation index can be used to determine whether the pixel points in the image are vegetation points.
  • the vegetation binary image can mean that any pixel in the image can only represent vegetation or non-vegetation.
  • the pixel points in the regional image are vegetation points, and different gray values are assigned to the vegetation points and other pixels to generate a vegetation binary map.
  • Step S106 Determine the connected domain of the crop from the connected vegetation of the first binarized image, and generate a second binary image based on the connected domain of the crop;
  • the vegetation connected domain can be a collection of pixels with connected gray values representing vegetation. Filtering crop connected domains in vegetation connected domains can be achieved by filtering according to the area of connected domains, filtering according to shapes, filtering according to shapes, and filtering according to textures.
  • the connected domains of crops can be filtered according to the area of the connected domains. Specifically, the area of all vegetation connected areas can be counted, and the area range of the crop connected areas can be determined through the average, mode, and dense interval of the area of the vegetation connected areas, or the area range of the crop connected areas can be artificially set.
  • the vegetation connected domains whose area is significantly smaller or larger than the area range of the crop connected domains are deleted, and the retained vegetation connected domains are the crop connected domains.
  • Step S108 for each planting row area in the second binarized image, determine the crop distance between any two adjacent crop connected areas in the planting row area, and/or the end crop connected areas in the planting row area The critical distance from the border of the planting row area, where the end crop connected domain is adjacent to only one crop connected domain in the planting row area;
  • the crop connection domain when the crop connection domain is located at the two ends of the planting row area, it is necessary to determine whether the distance between the crop connection domain (that is, the end crop connection domain) and the nearest crop boundary is greater than the defect. Seedling threshold distance.
  • step S110 it is determined whether the planting row area is in a crop missing state according to the crop spacing or the critical spacing.
  • the planting row area can be the area where a row of crops is located, and the width of the planting row area is greater than the maximum width of the crop connection area, and the crops in the planting row area are planted along the main direction of the planting row.
  • Fig. 1b provides a schematic diagram of a planting row area. As shown in Fig. 1b, the area between every two adjacent straight lines is the planting row area.
  • the cropping row area is divided along the main direction of the planting row in the crop binary image.
  • the advantage of this setting is that the crop connected domains in the crop binary map are divided into the planting row area according to the main direction of the planting row, so that the crop connected domains in the planting row area can be arranged in a row along the main direction of the planting row, which is convenient for the planting row area. Mark the lack of seedlings in the connected domains of the crops.
  • the critical distance it is necessary to obtain a crop boundary that matches the second binarized image; and then obtain an end crop boundary that matches the connected domain of the end crop from the crop boundary; Secondly, along the main planting row direction of the planting row area, calculate the critical distance between the center of mass of the connected area of the end crop and the boundary of the end crop; finally, when the critical distance is greater than the preset threshold, it is determined that the planting row area is in a crop missing state.
  • the area image of the target area is obtained, and secondly, the vegetation area is extracted from the area image, and the first binarized image corresponding to the vegetation area is generated, and then, from the first binarized image Determine the crop connected domain in the vegetation connected domain, and generate a second binarized image based on the crop connected domain.
  • the planting row area is determined to be in the crop missing state, so as to achieve the technical effect of timely, efficient and accurate identification of missing crops, thereby solving the manual detection of missing crops.
  • the above-mentioned preset threshold can be calculated as follows: Count the distance between adjacent crop connected areas in each planting row area by statistical screening method, obtain the mode or average, and calculate the mode or average by 1.6 times or 2 times The value of is used as the seedling shortage threshold distance, but this embodiment does not limit the calculation method of the seedling shortage threshold distance.
  • the planting row area when determining the distance between any two adjacent crop connected regions in the planting row area, first, the planting row area can be bidirectionally detected according to a preset threshold, wherein the two-way The detection consists of comparing the distance between the connected areas of each crop in the planting row area with a preset threshold according to one direction, and then compare the distance between the connected areas of each crop with the preset threshold according to the opposite direction of the direction; When the distance is greater than the preset threshold, it is determined that the planting row area is in a crop missing state. It should be noted that for any two adjacent crop connected domains, the detection result in any direction in the bidirectional detection result indicates any two adjacent When the distance between connected areas of crops is greater than a preset threshold, it is determined that the planting row area is in a crop missing state.
  • the main direction of the crop planting rows in the target area may be determined first, and secondly, based on each planting in the target area
  • the length of the row in the main direction generates the circumscribed rectangle of the second binary image, and the width direction of the rectangle is the ordinate and the length direction is the abscissa to establish a coordinate system, as shown in Figure 2; the function of the circumscribed rectangle is to make the crop
  • the connected domains of the crops in the binary graph are displayed along the main direction of the planting row, which is convenient for marking the lack of seedlings.
  • the number of non-vegetation pixels in the main direction can also be determined, and based on the number of non-vegetation pixels, a curve 30 indicating the distribution of the number of non-vegetation pixels is generated in the coordinate system, according to the curve
  • the peak apex of 30 and the straight line 32 in the set of straight lines corresponding to the main direction are cut into the second binarized image to obtain the planting row area, where the straight line 32 in the straight line set is used to indicate between adjacent planting rows.
  • the above curve can be a cumulative curve.
  • the above process can also be expressed as the following implementation: generate the circumscribed rectangle of the crop binary image according to the crop binary image and the main direction of the planting row; where the height direction of the circumscribed rectangle is the planting row main Direction; take the height direction (or width direction) of the circumscribed rectangle as the ordinate direction, establish a coordinate system, and project the crop binary map into the coordinate system; count the individual pixels in the non-crop connected domain in the ordinate direction According to the cumulative curve, at least one planting row area is obtained.
  • the abscissa of each peak apex in the accumulation curve when acquiring at least one planting row area, it can be achieved through the following process: acquiring the abscissa of each peak apex in the accumulation curve, and generating a straight line along the main direction of the planting row; in the crop binary image (ie, the second binary image ), the area between two adjacent straight lines is taken as the planting row area; after obtaining the abscissa of each peak apex in the cumulative curve, the abscissa of a target peak apex is obtained in turn as the current processing abscissa, and the current Process the neighborhood abscissa set associated with the abscissa; if the abscissa of the target neighborhood in the neighborhood abscissa set meets the critical point condition, replace the abscissa of the target peak apex with the abscissa of the target neighborhood; return to execution to obtain a target in turn
  • the critical point condition can be that the pixels in the planting row direction corresponding to the abscissa of the target neighborhood are all pixels in the non-crop connected domain, or the number of pixels in the non-crop connected domain far exceeds the number of pixels in the crop connected domain .
  • the abscissa of the target neighborhood exists in the abscissa set of the neighborhood, and the pixels in the corresponding planting row direction are all pixels in the non-crop connected domain, then the abscissa of the target neighborhood is The abscissa corresponding to the true peak apex before the cumulative curve offset, so the abscissa of the target peak apex is replaced with the abscissa of the target neighborhood.
  • the planting row is identified according to the angle value of each straight line with respect to the reference direction.
  • the main direction may include the following processes: mapping the pixel points in the connected domains of each crop to the Hough space of the polar coordinate system to obtain the crop point mapping result; obtaining the straight line detection fed back by the Hough space accumulator for the crop point mapping result As a result, statistical analysis is performed on the straight line detection result to obtain the main planting row direction of the farmland; wherein the straight line detection result includes a target number of straight lines and a straight line angle corresponding to each straight line.
  • the Hough space of the polar coordinate system may be a parameter space obtained after the Hough transform of the Cartesian coordinate system.
  • the crop point mapping result may be the result obtained in the Hough space after the pixel points in the connected domain of each crop are Hough transformed, and the crop point mapping result may be a straight line.
  • the Hough space accumulator can be used to count the straight line angles corresponding to the mapping results of each crop point.
  • the straight line angle may be the angle between the straight line and the viewing angle of the image. In a specific example, the straight line angle may be the angle between the width direction of the crop binary image and the straight line.
  • determining the main direction of crop planting rows in the target area can perform Hough transform on the second binarized image, and output the second binarized image through a Hough space accumulator
  • the main direction of the crop planting row is determined based on the angle parameter in the polar coordinate system parameter pair list, where the polar coordinate system parameter pair list includes at least one polar coordinate system parameter pair ,
  • Each polar coordinate system parameter alignment includes angle and radius.
  • the angle with the most occurrences in the polar coordinate parameter pair list is used as the main direction of the crop planting row; or, when the number of polar coordinate parameter pairs in the polar coordinate system parameter pair list is less than the predetermined value, the polar coordinate angle at the first position in the polar coordinate system parameter pair list is taken as the main direction of the crop planting row.
  • the curve 40 before cutting the second binarized image to obtain the planting row area according to the peak apex of the curve and the straight line in the set of straight lines corresponding to the main direction, as shown in FIG. 4, the curve 40 can be adjusted.
  • the smoothing methods include, but are not limited to, denoising after moving average, denoising after LOWESS smoothing, denoising after Univariate Spline fitting, and denoising after Savitzky_Golay Filter smoothing. Denoising situations include, but are not limited to, correcting the negative value of Savitzky_Golay Filter after smoothing, etc.
  • the smoothed curve includes non-vegetation pixels, offset the set of straight lines so that the target curve does not contain non-vegetation pixels.
  • Vegetation pixels specifically, if necessary, the straight line set can be shifted left and right within a certain neighborhood range, and the trigger condition of the shift may be that the original accumulation curve in the neighborhood has a horizontal axis point with a cumulative value of zero.
  • the reason for the offset is that it is possible that after a peak point is smoothed, the distance is not the original and the true zero point is offset. Then its translation is limited.
  • Set an adjacent distance threshold neighborhborhood). Search for the point where the cumulative value of the cumulative curve before smoothing is zero in the neighborhood of the peak.
  • the boundary line is missing at the left and right, the boundary line can be added.
  • the second binarized image may be extracted Sample data, where the sample data includes: crop connected domains; mark the crop connected domains in the sample data to obtain the crop label; train the crop recognition model based on the crop connected domains and the labels assigned to the crop connected domains to obtain
  • the above-mentioned crop recognition model can recognize based on the connected domains of the crops, so as to realize the purpose of identifying missing crops.
  • the crops can be manually labeled, or deep learning algorithms can be used to label to generate crop labels, where the crop labels can be irregular graphic connected domains, Gaussian circles, rectangular box collections, or polygons, etc. Conform to the shape of the input of the network.
  • the lack of seedling marking may be a marking performed between two adjacent connected domains where the lack of seedlings is detected, or between two adjacent connected domains.
  • two adjacent connected domains where a lack of seedlings are detected can be connected to a line segment as a mark of lack of seedlings.
  • the connected areas of crops at any other location can be supplemented according to the ideal adjacent distance.
  • the ideal adjacent distance can be based on each planting The average value or mode of the distance between adjacent crop connected areas in the row area is obtained.
  • the embodiments of the application do not limit the method and specific process of marking the lack of seedlings.
  • the lack of seedlings is marked according to the distance between the connected areas of adjacent crops in each planting row area.
  • the advantage of this setting is that it is judged whether there is a lack of seedlings according to the adjacent distance. For crops that are not strictly planted in a straight line, the lack of seedlings can also be marked, and the scope of application is wider.
  • the vegetation area is extracted from the regional image and the first binarized image corresponding to the vegetation area is generated.
  • the greenness index of each pixel in the regional image can be determined first; for each pixel Point, compare the magnitude of the greenness index and the greenness threshold; determine whether the pixel is a pixel in the vegetation area according to the comparison result; count the pixels in the vegetation area, and determine the first binary image based on the statistical result.
  • the vegetation is separated by the threshold Area, if the super green index of a pixel is greater than or equal to the threshold, it is a non-zero value, and if the super green index of a pixel is less than the threshold, the gray value of the non-zero value greater than or equal to the threshold is set to 255.
  • the gray value of the zero value is set to 0 to generate the first binary image
  • the threshold can be artificially set according to the planting area, type, shape, size, and planting experience of the crop, or it can be Obtained by the Ostu method.
  • the crop connected domain is determined from the vegetation connected domain of the first binarized image, and the second binarized image is generated based on the crop connected domain.
  • the first binarized image is obtained The area of each vegetation connected domain; for each vegetation connected domain, determine whether the area of the vegetation connected domain belongs to the preset value range, if it belongs to the preset value range, determine the vegetation connected domain as a crop connected domain; based on the determined crop
  • the connected domain generates a binary image.
  • methods for screening vegetation connected domains include but are not limited to the following methods: artificially set threshold method, statistical screening method, shape screening method, texture screening method, etc.
  • the vegetation connected domain is filtered by the shape screening method. For example, if a green truck is parked at the edge of the field, and the crop is an apple tree with an approximately circular shape, geometric analysis can be used to determine whether the vehicle is not. It belongs to the vegetation connected domain.
  • the statistical screening method is used to count the area of all connected domains, and the crop area range is determined by methods such as mode and dense interval, and the vegetation connected domains within the area range are screened.
  • the area interval of the vegetation connected domains is [20, 50] (unit/m2), the vegetation connected domains that fit this area range are left.
  • the size and mode of the vegetation connected domains are 30 square meters, then the connected domains with an area of 30 square meters can be reserved.
  • the above screening methods can be used in combination. For example, if a car parked at the edge of the field meets the super green index of the vegetation connected domain and the shape is similar to the shape of the crop, it can be used.
  • the area interval is further combined to determine whether the above-mentioned vehicle is a vegetation connected domain.
  • the distance set between any two adjacent crop connected areas in the target area can be counted; the average of all distances in the distance set can be calculated
  • the preset threshold is determined based on the average value, where the distance is the distance between the centroids of the adjacent crop connected domains; the preset threshold input by the user may also be directly accepted.
  • the preset threshold value is determined based on the above average value, and the average value may be determined as the preset threshold value; the method for determining the preset threshold value may also be: Calculate the radius of each crop connected area in the target area to obtain the crop For the average radius of the connected domain, the preset threshold is determined based on the average radius and the above average value. Specifically, the result obtained by summing the average radius and the average value may be used as the preset threshold.
  • the position in the planting row area in the crop missing state can be determined; and the position in the crop missing state is generated to indicate the crop missing state. Mark.
  • generating a mark for indicating the missing state of the crop at the position in the missing state of the crop includes: taking the position in the missing state as the center, and taking the average radius of the connected area of the crop as the radius, Generate a circular mark, and use the circular mark as a mark to indicate the missing state of the crop; or take the position in the missing state as the center and generate a rectangular mark with the side length N times the average radius of the connected domain of the crop, and The rectangular mark is used as a mark for indicating the missing state of the crop, in order to prevent the generated rectangular mark from overlapping the border of the adjacent crop connected domain or cover the adjacent crop connected domain, where 0 ⁇ N ⁇ 2.
  • the extreme position may refer to the two ends of the planting row area
  • the end crop boundary may refer to the crop boundary closest to the connected area of the end crop.
  • acquiring an area image of the target area may be receiving an area image of the target area taken by a drone.
  • Fig. 5 is a device for detecting crop absence in a target area according to an embodiment of the present application. As shown in Fig. 5, the device includes:
  • the obtaining module 50 is configured to obtain an area image of the target area
  • the first generating module 52 is configured to extract the vegetation area from the area image, and generate a first binary image corresponding to the vegetation area;
  • the second generation module 54 is configured to determine the connected domain of the crop from the vegetation connected domain of the first binary image, and generate a second binary image based on the connected domain of the crop;
  • the first determining module 56 is configured to determine, for each planting row area in the second binarized image, the crop spacing between any two adjacent crop connected areas in the planting row area, and/or the planting row area The critical distance between the end crop connected domain and the border of the planting row area, where the end crop connected domain is adjacent to only one crop connected domain in the planting row area;
  • the second determining module 58 is configured to determine whether the planting row area is in a crop missing state according to the crop spacing or the critical spacing.
  • the device for detecting the absence of crops in the target area includes: an acquiring module 50 configured to acquire an area image of the target area, and a first generating module 52 configured to extract a vegetation area from the area image and generate a first corresponding to the vegetation area. Binarized image; a second generation module 54 configured to determine the crop connected domain from the vegetation connected domain of the first binary image, and generate a second binarized image based on the crop connected domain; the first determining module 56, It is set to determine the distance between any two adjacent crop connected domains in the planting row area for each planting row area in the second binarized image; the second determining module 58 is set to be set when the distance is greater than a preset threshold When determining that the planting row area is in a crop missing state, the purpose of identifying missing crops is achieved, thereby realizing the technical effect of timely, efficient and accurate identification of missing crops, thereby solving the problem of manual detection of missing crops, which requires a lot of human participation and time-consuming Technical problems that are laborious and prone to human error.
  • FIG. 6 is an unmanned aerial vehicle according to an embodiment of the present application.
  • the unmanned aerial vehicle includes: an image acquisition device 60 configured to acquire an area image of a target area; a processor 62 configured to obtain an image from the area Extract the vegetation area from the, and generate the first binarized image corresponding to the vegetation area; determine the crop connected domain from the vegetation connected domain of the first binarized image, and generate the second binarized image based on the crop connected domain; For each planting row area in the second binary image, determine the crop spacing between any two adjacent crop connected areas in the planting row area, and/or the end crop connected areas and the planting row area in the planting row area The critical distance between the borders of, wherein the end crop connected domain is adjacent to only one crop connected domain in the planting row area; whether the planting row area is in a crop missing state is determined according to the crop distance or the critical distance.
  • Fig. 7 is a method for calculating the number of missing crops according to an embodiment of the present application. The method includes the following steps:
  • Step S702 Obtain an area image of the target area
  • Step S704 extract the vegetation area from the area image, and generate a first binary image corresponding to the vegetation area;
  • Step S706 Determine the crop connected domain from the vegetation connected domain of the first binarized image, and generate a second binarized image based on the crop connected domain;
  • Step S708 For each planting row area in the second binarized image, determine the distance between any two adjacent crop connected regions in the planting row area;
  • Step S710 when the distance is greater than a preset threshold, it is determined that the planting row area is in a crop missing state, and a crop missing mark is generated;
  • step S712 the position of the missing seedlings is identified according to the missing mark, and the number of missing crops is calculated.
  • the area image of the target area is obtained; secondly, the vegetation area is extracted from the area image, and the first binarized image corresponding to the vegetation area is generated, and then the first binarized image is obtained from the first binarized image.
  • Determine the crop connected domain in the vegetation connected domain and generate a second binarized image based on the crop connected domain, and then for each planting row area in the second binarized image, determine any two adjacent planting row areas
  • the distance between the connected domains of crops is greater than a preset threshold, it is determined that the planting row area is in a crop missing state, and a crop missing mark is generated.
  • the position of missing seedlings is identified, and the number of missing crops is calculated.
  • Fig. 8 is a device for calculating the number of missing crops according to an embodiment of the present application.
  • the device includes: an acquisition module 80, configured to acquire an area image of a target area; an identification module 82, configured to extract a vegetation area from the area image, and Generate the first binarized image corresponding to the vegetation area. For each planting row area in the second binarized image, determine the crop spacing between any two adjacent crop connected domains in the planting row area, and/ Or the critical distance between the end crop connected domain in the planting row area and the border of the planting row area; wherein, the end crop connected domain is adjacent to only one crop connected domain in the planting row area; the calculation module 84 is set according to The crop spacing or critical spacing determines whether the planting row area is in a crop missing state.
  • Fig. 9 is another crop missing detection method according to an embodiment of the present application. As shown in Fig. 9, the method includes the following steps:
  • Step S902 Obtain an area image of the target area; the foregoing area image includes but is not limited to an image of a farmland, for example, it may be a top view of the farmland.
  • Step S904 Input the region image to the crop recognition model for analysis to obtain each crop connected domain in the target area, and generate a binarized image based on the crop connected domain.
  • the crop recognition model is trained based on multiple sets of data.
  • Each group of data in the group data includes a sample image and a label used to mark the connected domains in the sample image;
  • Step S906 Determine the crop distance between any two adjacent crop connected areas in each planting row area in the binarized image, and/or the convenient distance between the end crop connected areas in the planting row area and the planting row area Critical spacing of, in which, the end crop communication domain has only one adjacent crop communication domain in the planting row area;
  • step S908 it is determined whether the planting row area is in a crop missing state according to the crop spacing or the critical spacing.
  • the area image of the target area is obtained; secondly, the area image is input into the crop recognition model for analysis, and each crop connected domain in the target area is obtained, and the binary image is generated based on the crop connected domain , Where the crop recognition model is trained based on multiple sets of data.
  • Each set of data in the multiple sets of data includes a sample image and a label for marking the connected domains in the sample image; finally, each planting row area is determined When the distance between any two adjacent crop connected domains in, when the distance is greater than the preset threshold, it is determined that the planting row area is in a crop missing state, thereby realizing the technical effect of timely, efficient and accurate identification of missing crops, and then solving The manual detection of missing crops requires a lot of human participation, time-consuming and labor-intensive, and is prone to technical problems such as human error.
  • determining the distance between any two adjacent crop connected regions in each planting row area includes: performing bidirectional detection on the planting row area according to a preset threshold, wherein the two-way The detection is to compare the distance between the connected areas of each crop in the planting row area with a preset threshold according to one direction, and then compare the distance between the connected areas of each crop with the preset threshold in the opposite direction of the direction; When it is greater than the preset threshold, it is determined that the planting row area is in the crop missing state, including: for any two adjacent crop connected domains, the detection result in any direction in the bidirectional detection result indicates one of any two adjacent crop connected domains When the distance between the two is greater than the preset threshold, it is determined that the planting row area is in a crop missing state.
  • the method further includes: determining the main direction of the crop planting rows in the target area; The length of each planting row in the area in the main direction generates the circumscribed rectangle of the binarized image; the width of the rectangle is the ordinate and the length is the abscissa to establish a coordinate system; determine the number of non-vegetation pixels in the main direction, and Based on the number of non-vegetation pixels, a curve indicating the distribution of the number of non-vegetation pixels is generated in the coordinate system; the second binarized image is cut according to the peak apex of the curve and the line in the line set corresponding to the main direction , The planting row area is obtained, where the straight line in the straight line set is used to indicate the set of position points where the non-vegetation pixels are located between adjacent planting rows.
  • Fig. 10 is a schematic diagram of an optional crop missing detection principle of the present application. As shown in Fig. 10, the process mainly includes the following steps:
  • a large number of photos of farmland are taken overhead to form a collection of farmland photos.
  • the algorithm is used to extract the area where the vegetation is located in the photo, and all the vegetation connected domains are screened, leaving the connected domains of the crops, generating a binarized map of the crops, and then generating Crop labels, or directly generate crop labels through manual labeling, to obtain training data sets, test data sets, and verification data sets stored in pairs of farmland photos and generated crop labels;
  • train the training data set through the depth network Obtain the deep learning model, the verification set is used to adjust the hyperparameters, the parameters corresponding to the optimal model are selected, and the test set is used to measure the optimal performance; finally, the crop segmentation model is obtained.
  • the crop segmentation model is used to identify the image to be executed, obtain the crop segmentation map, and use the algorithm to obtain the main direction of the crop planting row.
  • the above algorithm includes but not limited to the method of using Hough to detect straight lines.
  • the crop segmentation map is first
  • the corresponding binarization graph is converted to the polar coordinate system Hough space, and the polar coordinate angle ⁇ and radius ⁇ of the front line are returned through the Hough space accumulator.
  • the list data is statistically analyzed to obtain the main planting row. direction.
  • Statistical analysis methods include, but are not limited to, selecting the angle with the most occurrences as the main direction of the planting row, and selecting the first angle as the main direction of the planting row when there are fewer returned results, and then generating a binary map with the main direction as the height
  • the circumscribed rectangle of, the width direction (main direction) of the circumscribed rectangle is the ordinate, the length direction is the abscissa, and the number of non-vegetation pixels in the main direction is accumulated to obtain the cumulative curve, because in fact the direction of the planting row (main direction) is not necessarily It is perpendicular to the viewing angle of the picture.
  • the purpose of the circumscribed rectangle is to add zero pixels along the main direction to obtain the cumulative curve for waveform analysis.
  • the generation of the missing marker can be a point where the distance between the centroid on the line direction of the centroid of the current connected domain is the average adjacent distance, a circle with an average radius, or a rectangle with a side length of N times the radius, where 0 ⁇ N ⁇ 2.
  • any crop missing detection method can be used to determine the lack of seedlings; then, according to the lack of seedlings, a recommended route for agricultural machinery operations is generated, and the lack of seedlings can be replanted.
  • the crops can be The detection method determines the missing seedling area, and then the agricultural machinery can directly perform replanting operations on these several missing seedling areas without having to check the entire crop area one by one before replanting.
  • the process of determining the lack of seedling area includes: processing the area image of the target area based on the method for detecting the lack of crops according to any one of claims 1 to 20, and determining the location of the target area. Whether the included planting row area is in the crop missing state; when it is determined that the planting row area is in the crop missing state, a crop missing mark is generated; according to the image position of the crop missing mark in the regional image binary map, and the match with the regional image binary map Geographical location information, determine the geographic location information corresponding to the missing markers of each crop, to obtain the seedling-deficient area.
  • an operation route planning method Specifically, any crop missing detection method is adopted to determine the area of lack of seedlings; the operation route is determined according to the area of lack of seedlings, for example, when the agricultural machinery is in When performing operations, it is possible to first identify the area with missing seedlings based on the detection method of crop missing, and then the agricultural machinery can plan a route to avoid the area with missing seedlings based on these several areas with missing seedlings.
  • an operation method is also provided. Specifically, the operation equipment first executes any kind of crop loss The detection method determines the lack of seedling area; then the operating equipment marks the lack of seedling area to determine the position of the lack of seedlings; finally, when the location of the operating equipment is located at the position of lack of seedlings, the control equipment stops the operation at the position of lack of seedlings.
  • the above-mentioned equipment includes but is not limited to the following types: spraying equipment, spreading equipment, and harvesting equipment.
  • a spraying device when a spraying device is spraying pesticides, it can obtain the lack of seedling area according to the detection method of crop lack. When the spraying device falls into the range of the lack of seedling area, the spraying operation is stopped; for another example, when the spreading device is spreading solid fertilizer , When the spreading equipment falls into the range of the lack of seedlings, stop the spreading operation. For another example, when harvesting equipment is harvesting crops, it can retract the harvesting device by itself and suspend the harvesting operation when it travels to an area lacking seedlings.
  • a yield measurement method is also provided. Specifically: first, the above-mentioned crop missing detection method can be used to determine the crop lacking area; Then, the area of the non-deficient seedling area is determined according to the lack of seedlings; finally, the total yield of the crop area is determined by the product of the yield per unit area of the non-deficient area and the total area of the non-deficient area, for example, the total area of a piece of farmland is The yield per unit area of 1 hectare and the non-missing area is 1 ton.
  • the crop missing detection method After the above-mentioned crop missing detection method, several missing areas in this farmland have been identified, and then the area of the missing seedlings area is 0.1 hectares.
  • the size of the area of the lack of seedlings is the total area of the farmland minus the area of the lack of seedlings equal to 0.9 hectares, and then multiplied by the above unit yield of 1 ton, the crop yield of the non-missing area is 0.9 tons, that is, the crop yield of this farmland.
  • the output is 0.9 tons.
  • a non-volatile storage medium includes a stored program, wherein, when the program is running, the device where the non-volatile storage medium is located is controlled to execute arbitrary A method for detecting missing crops, or any method for calculating the number of missing crops, or any operation path planning method, or any replanting method, or any operation method, or any yield measurement method .
  • the above-mentioned non-volatile storage medium is used to store program instructions for performing the following functions, so as to realize the following functions:
  • Obtain the regional image of the target area extract the vegetation area from the regional image, and generate the first binary image corresponding to the vegetation area; determine the crop connected domain from the vegetation connected domain of the first binary image, and based on the crop connection Domain generates a second binarized image; for each planting row area in the second binarized image, determine the crop spacing between any two adjacent crop connected domains in the planting row area, and/or the planting row area.
  • a processor wherein the processor is used to run a program, wherein, when the program is running, any method for detecting missing crops or calculating the number of missing crops is executed.
  • the above-mentioned processor is used to call program instructions in the memory to realize the following functions:
  • Obtain the regional image of the target area extract the vegetation area from the regional image, and generate the first binary image corresponding to the vegetation area; determine the crop connected domain from the vegetation connected domain of the first binary image, and based on the crop connection Domain generates a second binarized image; for each planting row area in the second binarized image, determine the crop spacing between any two adjacent crop connected domains in the planting row area, and/or the planting row area.
  • the disclosed technical content can be implemented in other ways.
  • the device embodiments described above are merely illustrative.
  • the division of the units may be a logical function division.
  • multiple units or components may be combined or may be Integrate into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, units or modules, and may be in electrical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.
  • the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium.
  • the technical solution of the present application essentially or the part that contributes to the existing technology or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , Including several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in each embodiment of the present application.
  • the aforementioned storage media include: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

一种作物缺失的检测方法及检测装置。其中，所述方法包括：获取目标区域的区域图像（S102）；从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像（S104）；从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像（S106）；对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距，其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻（S108）；根据作物间距或临界间距确定种植行区域是否处于作物缺失状态（S110）。

Description

作物缺失的检测方法及检测装置
本申请要求于2020年6月12日提交中国专利局、申请号为202010537090.3、申请名称为“一种缺苗标记方法、装置、计算机设备和存储介质”的中国专利申请的优先权，其全部内容通过引用结合中本申请中。
本申请要求于2020年6月12日提交中国专利局、申请号为202010538238.5、申请名称为“作物缺失的检测方法及检测装置”的中国专利申请的优先权，其全部内容通过引用结合中本申请中。
技术领域
本申请涉及农作物识别领域，具体而言，涉及一种作物缺失的检测方法及检测装置。
背景技术
在现代农业种植中，作物缺失的数目，不仅可以作为评价作物的种子质量、作业机器精度、农田环境等重要的参考依据，同时也可以及时发现缺苗并做出移栽等补救措施弥补产量损失，因此及时、高效、准确地检测出缺失作物具有非常重要的意义。
目前检测缺失的作物方法主要依靠人工测量数目，通过人工检测缺失作物，需要大量人力参与、耗时费力，且容易出现人为失误。
针对上述的问题，目前尚未提出有效的解决方案。
发明内容
本申请实施例提供了一种作物缺失的检测方法及检测装置，以至少解决人工检测缺失作物，需要大量人力参与、耗时费力，且容易出现人为失误的的技术问题。
根据本申请实施例的一个方面，提供了一种作物缺失的检测方法，包括：获取目标区域的区域图像；从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距，其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
可选地，根据作物间距确定种植行区域是否处于作物缺失状态，包括：依据预设阈值对种植行区域进行双向检测，其中，该双向检测为依据一个方向对种植行区域中的各个作物连通域之间的距离与预设阈值进行比较后，再按照方向的反方向对各个作物连通域之间的距离与预设阈值进行比较；距离大于预设阈值时，确定种植行区域处于作物缺失状态，包括：对于任意两个相邻的作物连通域，在双向检测结果中任意一个方向的检测结果指示任意两个相邻的作物连通域之间的距离大于预设阈值时，确定 种植行区域处于作物缺失状态。
可选地，根据临界间距确定种植行区域是否处于作物缺失状态，包括：获取与第二二值化图像匹配的作物边界；从作物边界中获取与端部作物连通域匹配的端部作物边界；沿种植行区域的种植行主方向，计算端部作物连通域的质心与端部作物边界之间的临界间距；在临界间距大于预设阈值时，确定种植行区域处于作物缺失状态。
可选地，第二二值化图像中的每个种植行区域的确定过程包括：确定目标区域中作物种植行的主方向；基于目标区域中各个种植行在主方向的长度生成第一二值化图像的外接矩形；并以矩形的宽度为纵坐标，长度方向为横坐标，建立坐标系；确定主方向上非植被像素的个数，并基于非植被像素的个数，在坐标系中生成用于指示非植被像素的个数分布情况的曲线；依据曲线的波峰顶点、主方向对应的直线集合中的直线，切割第二二值化图像，得到种植行区域，其中，直线集合中的直线用于指示相邻种植行之间，非植被像素所在的位置点集合。
可选地，确定目标区域中作物种植行的主方向，包括：对第二二值化图像进行霍夫变换，并通过霍夫空间累加器输出第二二值化图像在霍夫空间中的极坐标系参数对列表，基于极坐标系参数对列表中的角度参数确定作物种植行的主方向，其中，极坐标系参数对列表中的包括至少一个极坐标系参数对，每个极坐标系参数对中均包括角度和半径。
可选地，通过霍夫空间累加器输出第二二值化图像在霍夫空间中的极坐标系参数对列表，基于极坐标系参数对列表中的角度参数确定作物种植行的主方向，包括：在极坐标系参数对列表中的极坐标参数对的数量大于预定值时，将极坐标参数对列表中出现次数最多的角度作为作物种植行的主方向；或者，在极坐标系参数对列表中的极坐标参数对的数量小于预定值时，将极坐标系参数对列表中位于首位的极坐标角度作为作物种植行的主方向。
可选地，依据曲线的波峰顶点、主方向对应的直线集合中的直线，切割第二二值化图像得到种植行区域之前，方法还包括：对曲线进行平滑处理，得到平滑处理后的目标曲线；在进行平滑处理之后的曲线中包括非植被像素时，对直线集合进行偏移，以使得目标曲线中不包含非植被像素。
可选地，依据曲线的波峰顶点、主方向对应的直线集合中的直线，切割第二二值化图像，得到种植行区域，包括：获取曲线中的各波峰顶点的横坐标，沿种植行主方向生成直线；在第二二值图中，将相邻两条直线之间的区域作为种植行区域。
可选地，在获取曲线中的各波峰顶点的横坐标之后，还包括：依次获取一个目标波峰顶点的横坐标作为当前处理横坐标，并获取当前处理横坐标关联的邻域横坐标集合；如果邻域横坐标集合中目标邻域横坐标满足临界点条件，则将目标波峰顶点的横坐标替换为目标邻域横坐标；返回执行依次获取一个目标波峰顶点的横坐标作为当前处理横坐标的操作，直至完成对全部波峰顶点的横坐标的处理。
可选地，从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像之后，方法还包括：从第二二值化图像中提取样本数据，其中，该样本数据中包括：作物连通域；对样本数据中的作物连通域进行标记，得到作物标签；基于作物连通域和为作物连通域分配的标签对作物识别模型进行训练，得到训练 后的作物识别模型。
可选地，从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像，包括：确定区域图像中各个像素点的绿度指数；对于每个像素点，比较绿度指数和绿度阈值的大小；依据比较结果确定像素点是否为植被区域中的像素点；统计属于植被区域中的像素点，并基于统计结果确定第一二值化图像。
可选地，从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像，包括：获取第一二值化图像中各个植被连通域的面积；对于每个植被连通域，确定植被连通域的面积是否属于预设取值范围，如果属于预设取值范围，则确定植被连通域为作物连通域；基于确定的作物连通域生成二值化图像。
可选地，预设阈值通过以下任一种方式获取：第一种方式：统计目标区域中任意两个相邻的作物连通区域质心之间的距离集合；计算距离集合中的所有距离的平均值，并基于平均值确定预设阈值，其中距离为相邻的作物连通域质心之间的距离；第二种方式：接收用户输入的预设阈值。
可选地，基于平均值确定预设阈值，包括：将平均值确定为预设阈值；或者，统计目标区域中各个作物连通区域的半径，得到作物连通域的平均半径，基于平均半径和平均值确定预设阈值。
可选地，确定种植行区域处于作物缺失状态之后，方法还包括：确定种植行区域中处于作物缺失状态的位置；并在处于作物缺失状态的位置处生成用于指示作物缺失状态的标记。
可选地，在处于作物缺失状态的位置处生成用于指示作物缺失状态的标记，包括：以处于缺失状态的位置为中心，以连通域的平均半径为半径，生成圆形标记，并将该圆形标记作为用于指示作物缺失状态的标记；或者以处于缺失状态的位置为中心，以连通域的平均半径的2倍为边长生成矩形标记，并将该矩形标记作为用于指示作物缺失状态的标记。
可选地，获取目标区域的区域图像，包括：接收无人机拍摄的目标区域的区域图像。
根据本申请实施例的另一方面，还提供了一种作物缺失的检测方法，其中，包括：获取目标区域的区域图像；将区域图像输入至作物识别模型进行分析，得到目标区域中的各个作物连通域，并基于作物连通域生成二值化图像，其中，作物识别模型为基于多组数据训练得到的，多组数据中的每组数据中均包括样本图像，以及用于标记样本图像中的连通域的标签；确定二值化图像中每个种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距；其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
可选地，根据作物间距确定种植行区域是否处于作物缺失状态，包括：依据预设阈值对种植行区域进行双向检测，其中，该双向检测为依据一个方向对种植行区域中的各个作物连通域之间的距离与预设阈值进行比较后，再按照方向的反方向对各个作物连通域之间的距离与预设阈值进行比较；对于任意两个相邻的作物连通域，在双向 检测结果中任意一个方向的检测结果指示任意两个相邻的作物连通域之间的作物间距大于预设阈值时，确定种植行区域处于作物缺失状态。
可选地，二值化图像中的每个种植行区域的确定过程包括：确定目标区域中作物种植行的主方向；基于目标区域中各个种植行在主方向的长度生成二值化图像的外接矩形；并以矩形的宽度为纵坐标，长度方向为横坐标，建立坐标系；确定主方向上非植被像素的个数，并基于非植被像素的个数，在坐标系中生成用于指示非植被像素的个数分布情况的曲线；依据曲线的波峰顶点、主方向对应的直线集合中的直线，切割第二二值化图像，得到种植行区域，其中，直线集合中的直线用于指示相邻种植行之间，非植被像素所在的位置点集合。
根据本申请的另一方面，还提供了一种缺失作物数的计算方法，包括：基于作物缺失的检测方法，对目标区域的区域图像进行处理，确定目标区域所包含的种植行区域是否处于作物缺失状态；确定种植行区域处于作物缺失状态时，生成作物缺失标记；依据作物缺失标记，计算缺失的作物数目。
根据本申请的另一方面，还提供了一种作业路线规划方法，包括：基于作物缺失的检测方法，确定缺苗区域；根据缺苗区域生成作业路线，其中，作业路线经过缺苗区域。
根据本申请的另一方面，还提供了一种补种方法，包括：基于作物缺失的检测方法，确定缺苗区域；根据缺苗区域生成作业路线；向作业设备发送作业路线，其中，作业路线用于指示作业设备对缺苗区域进行作物补种。
可选地，缺苗区域的确定过程包括：基于作物缺失的检测方法，对目标区域的区域图像进行处理，确定目标区域所包含的种植行区域是否处于作物缺失状态；确定种植行区域处于作物缺失状态时，生成作物缺失标记；根据作物缺失标记在区域图像二值图中的图像位置，以及与区域图像二值图匹配的地理位置信息，确定各作物缺失标记对应的地理位置信息，以得到缺苗区域。
可选地，根据缺苗区域生成补种作业路线，包括：根据各作物缺苗标记对应的地理位置信息，生成与种植区域匹配的补种作业路线。
根据本发明实施例的另一方面，还提供了一种作业方法，应用于作业设备，作业设别包括以下至少之一：喷洒设备、播撒设备和采收设备；方法包括：确定作业设备的当前位置是否位于目标区域的缺苗位置，如果作业设备位于缺苗位置，则停止在缺苗位置进行作业，其中，缺苗位置基于作物缺失的检测方法确定得到。
根据本发明实施例的另一方面，还提供了一种产量测算方法，包括：基于作物缺失的检测方法确定目标区域中的缺苗区域；根据缺苗区域确定目标区域中非缺苗区域的面积；基于非缺苗区域的单位面积产量与非缺苗区域的总面积确定目标区域的总产量。
根据本申请实施例的另一方面，还提供了一种目标区域中作物缺失的检测装置，包括：获取模块，设置为获取目标区域的区域图像；第一生成模块，设置为从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；第二生成模块，设置为从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二 值化图像；第一确定模块，设置为对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距，其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；第二确定模块，设置为根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
根据本申请实施例的另一方面，还提供了一种无人机，包括：图像采集装置，设置为获取目标区域的区域图像；处理器，设置为从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像，从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距，其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
根据本申请实施例的另一方面，还提供了一种缺失作物数的计算装置，包括：获取模块，设置为获取目标区域的区域图像；识别模块，设置为从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距；其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；计算模块，设置为根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
根据本申请实施例的另一方面，还提供了一种非易失性存储介质，该非易失性存储介质包括存储的程序，其中，在程序运行时控制非易失性存储介质所在设备执行任意一种作物缺失的检测方法，或任意一种缺失作物数的计算方法，或任意一种作业路径规划方法，或任意一种补种方法，或任意一种作业方法，或任意一种产量测算方法。
根据本申请实施例的另一方面，还提供了一种处理器，其中，处理器用于运行程序，其中，程序运行时执行任意一种作物缺失的检测方法，或任意一种缺失作物数的计算方法，或任意一种作业路径规划方法，或任意一种补种方法，或任意一种作业方法，或任意一种产量测算方法。
在本申请实施例中，采用基于深度学习算法识别区域图像的方式，通过获取目标区域的区域图像；从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距，其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；根据作物间距或临界间距确定种植行区域是否处于作物缺失状态，达到了识别缺失作物的目的，从而实现了及时、高效，准确地识别缺失作物的技术效果，进而解决了人工检测缺失作物，需要大量人力参与、耗时费力，且容易出现人为失误的技术问题。
附图说明
此处所说明的附图用来提供对本申请的进一步理解，构成本申请的一部分，本申请的示意性实施例及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：
图1a是根据本申请实施例的一种作物缺失的检测方法的流程示意图；
图1b是根据本申请实施例的一种种植行区域的示意图；
图2是根据本申请实施例的一种可选的以外接矩形宽度为纵坐标建立的坐标系示意图；
图3是根据本申请实施例的一种可选的曲线分割的框架示意图；
图4是根据本申请实施例的一种可选的平滑曲线的框架示意图；
图5是根据本申请实施例的一种作物缺失的检测装置的结构示意图；
图6是是根据本申请实施例的一种可选的无人机的结构示意图；
图7是根据本申请实施例的一种缺失作物数的计算方法的流程示意图；
图8是根据本申请实施例的一种缺失作物数的计算装置的结构示意图；
图9是根据本申请实施例的另一种作物缺失的检测方法的流程示意图；
图10是根据本申请实施例的一种可选的作物缺失的检测原理示意图。
具体实施方式
为了使本技术领域的人员更好地理解本申请方案，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分的实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都应当属于本申请保护的范围。
需要说明的是，本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
根据本申请实施例，提供了一种作物缺失的检测的方法实施例，需要说明的是，在附图的流程图示出的步骤可以在诸如一组计算机可执行指令的计算机系统中执行，并且，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤。
图1a是根据本申请实施例的作物缺失的检测方法，如图1a所示，该方法包括如下步骤：
步骤S102，获取目标区域的区域图像；
步骤S104，从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；
提取植被区域的实现方式包括但不限于：根据区域图像中包括的植被点，生成植被二值图，即第一二值化图像。具体地：
植被点表明图像中的像素点代表植被，可以通过颜色空间的转换、颜色指数以及植被指数等方式确定图像中的像素点是否为植被点。在一个具体的示例中，可以通过植被指数中的过绿指数(Excess Green Index，EXG)判断图像中的像素点是否为植被点。具体的，通过下述公式计算像素点的过绿指数：EXG＝2*Green-Red-Blue，Green、Red、Blue表示像素点的RGB指数经过归一化处理后得到的数值。如果像素点的过绿指数大于一定阈值，则判断为植被，否则为非植被。植被二值图可以指图像中的任何像素点只能代表植被或非植被。
在本发明实施例中，判断区域图像中的像素点是否为植被点，对植被点和其他像素点赋予不同的灰度值，生成植被二值图。
步骤S106，从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；
植被连通域可以为彼此连通的灰度值代表植被的像素点的集合。在植被连通域中筛选作物连通域，可以通过按照连通域面积筛选、按照形状筛选、按照形状筛选以及按照纹理筛选等方式实现。
在一个具体的示例中，可以按照连通域面积筛选作物连通域。具体的，可以统计所有植被连通域的面积，通过植被连通域面积的平均数、众数以及密集区间等确定作物连通域的面积范围，或人为设定作物连通域的面积范围。将面积明显小于或大于作物连通域的面积范围的植被连通域删除，保留的植被连通域即为作物连通域。
步骤S108，对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距，其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；
在本申请的实施例中，当作物连通域位于种植行区域两端处，需要判断该作物连通域(也就是端部作物联通域)与距离其最近的作物边界之间的距离，是否大于缺苗阈值距离。
步骤S110，根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
其中，种植行区域可以为一行作物所在的区域，种植行区域的宽度大于作物连通 域的最大宽度，种植行区域内的作物，沿种植行主方向种植。
示例性地，图1b提供了一种种植行区域的示意图，如图1b所示，每两条相邻直线之间的区域即为种植行区域。
在本申请的一些实施例中，根据农田图像获取作物二值图，并获取种植行主方向之后，在作物二值图内沿种植行主方向分割种植行区域。
这样设置的优点在于，将作物二值图内的作物连通域按照种植行主方向划分种植行区域，可以使种植行区域内的作物连通域沿种植行主方向呈一列排列，便于在种植行区域内对作物连通域进行缺苗标记。
可选地，根据临界间距确定种植行区域是否处于作物缺失状态，需要获取与第二二值化图像匹配的作物边界；然后从作物边界中获取与端部作物连通域匹配的端部作物边界；其次沿种植行区域的种植行主方向，计算端部作物连通域的质心与端部作物边界之间的临界间距；最后在临界间距大于预设阈值时，确定种植行区域处于作物缺失状态。
上述作物缺失的检测方法中，首先，获取目标区域的区域图像，其次，从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像，然后，从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像，最后，对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的距离；在距离大于预设阈值时，确定种植行区域处于作物缺失状态，从而实现了及时、高效，准确地识别缺失作物的技术效果，进而解决了人工检测缺失作物，需要大量人力参与、耗时费力，且容易出现人为失误的技术问题。
需要说明的是，小于或者等于预设阈值时，确定种植行区域不处于作物缺失状态。
上述预设阈值的计算方式可以为：通过统计筛选法统计各种植行区域内相邻作物连通域之间的距离，获取众数或平均数，将众数或平均数的计算1.6倍或2倍的值作为缺苗阈值距离，但本实施例不限制缺苗阈值距离的计算方式。
本申请一种可选的实施例中，确定种植行区域中任意两个相邻的作物连通域之间的距离时，首先，可以依据预设阈值对种植行区域进行双向检测，其中，该双向检测为依据一个方向对种植行区域中的各个作物连通域之间的距离与预设阈值进行比较后，再按照方向的反方向对各个作物连通域之间的距离与预设阈值进行比较；然后在距离大于预设阈值时，确定种植行区域处于作物缺失状态，需要说明的是：对于任意两个相邻的作物连通域，在双向检测结果中任意一个方向的检测结果指示任意两个相邻的作物连通域之间的距离大于预设阈值时，则确定种植行区域处于作物缺失状态。
需要说明的是，对于端点处缺失作物连通域的情况，在端点相邻的作物域以预设阈值进行双向检测时，其中一个方向必然为端点方向，端点方向在预设阈值范围内，若搜索不到作物连通域，则认为此端点处缺失作物连通域。
本申请一种可选的实施例中，从第一二值化图像的植被连通域中确定作物连通域之后，可以先确定目标区域中作物种植行的主方向，其次，基于目标区域中各个种植行在主方向的长度生成第二二值化图像的外接矩形，并以矩形的宽度方向为纵坐标，长度方向为横坐标，建立坐标系，如图2所示；外接矩形的作用在于使作物二值图中 的作物连通域沿种植行主方向陈列，便于进行缺苗标记。
如图3所示，还可以确定主方向上非植被像素的个数，并基于非植被像素的个数，在坐标系中生成用于指示非植被像素的个数分布情况的曲线30，依据曲线30的波峰顶点、主方向对应的直线集合中的直线32，切割第二二值化图像，得到所述种植行区域，其中，直线集合中的直线32用于指示相邻种植行之间，非植被像素所在的位置点集合。
上述曲线可以为累加曲线，此时，上述过程也可以表现为以下实现方式：根据作物二值图和种植行主方向生成作物二值图的外接矩形；其中，外接矩形的高度方向为种植行主方向；以外接矩形的高度方向(或称为宽度方向)为纵坐标方向，建立坐标系，并将作物二值图投影于坐标系中；统计纵坐标方向上非作物连通域内的像素点的个数，生成累加曲线；根据累加曲线，获取至少一个种植行区域。
其中，在获取至少一个种植行区域时，可以通过以下过程实现：获取累加曲线中的各波峰顶点的横坐标，沿种植行主方向生成直线；在作物二值图(即第二二值化图像)中，将相邻两条直线之间的区域作为种植行区域；在获取累加曲线中的各波峰顶点的横坐标之后，依次获取一个目标波峰顶点的横坐标作为当前处理横坐标，并获取当前处理横坐标关联的邻域横坐标集合；如果邻域横坐标集合中目标邻域横坐标满足临界点条件，则将目标波峰顶点的横坐标替换为目标邻域横坐标；返回执行依次获取一个目标波峰顶点的横坐标作为当前处理横坐标的操作，直至完成对全部波峰顶点的横坐标的处理。
临界点条件可以为，目标邻域横坐标对应的种植行方向上的像素点全部为非作物连通域内的像素点，或者，非作物连通域内的像素点的数量值远远超过作物连通域内的像素点。
在本申请的一些实施例中，如果邻域横坐标集合中存在目标邻域横坐标，其对应的种植行方向上的像素点全部为非作物连通域内的像素点，则说明目标邻域横坐标为累加曲线偏移之前真正的波峰顶点对应的横坐标，因此将目标波峰顶点的横坐标替换为目标邻域横坐标。
本申请一些实施例中，按照预设的直线检测算法，获取与所述作物二值图中的各作物连通域对应的至少一条直线，并根据各直线相对于参考方向的角度值，识别种植行主方向，可以包括以下过程：将所述各作物连通域内的像素点映射至极坐标系霍夫空间中，得到作物点映射结果；获取霍夫空间累加器针对所述作物点映射结果反馈的直线检测结果，对所述直线检测结果进行统计分析，得到所述农田的种植行主方向；其中，所述直线检测结果中包括目标数量的直线，以及与各直线对应的直线角度。
其中，极坐标系霍夫空间可以为对笛卡尔坐标系进行霍夫变换之后得到的参数空间。作物点映射结果可以为各作物连通域内的像素点经过霍夫变换之后在霍夫空间内得到的结果，作物点映射结果可以为一条直线。霍夫空间累加器可以用于统计各作物点映射结果对应的直线角度。直线角度可以为直线与图像的观察角度之间的夹角，在一个具体的示例中，直线角度可以为作物二值图的宽的方向与直线的夹角。
本申请一种可选的实施例中，确定目标区域中作物种植行的主方向，可以对对第二二值化图像进行霍夫变换，并通过霍夫空间累加器输出第二二值化图像在霍夫空间 中的极坐标系参数对列表，基于极坐标系参数对列表中的角度参数确定作物种植行的主方向，其中，极坐标系参数对列表中的包括至少一个极坐标系参数对，每个极坐标系参数对中均包括角度和半径。
本申请的一些实施例中，在极坐标系参数对列表中的极坐标参数对的数量大于预定值时，将极坐标参数对列表中出现次数最多的角度作为作物种植行的主方向；或者，在极坐标系参数对列表中的极坐标参数对的数量小于预定值时，将极坐标系参数对列表中位于首位的极坐标角度作为作物种植行的主方向。
本申请一种可选的实施例中，依据曲线的波峰顶点、主方向对应的直线集合中的直线，切割第二二值化图像得到种植行区域之前，如图4所示，可以对曲线40进行平滑处理，得到平滑处理后的目标曲线42，其中，平滑处理的方法包括但不限于移动平均后去噪、LOWESS平滑后去噪，Univariate Spline拟合后去噪、Savitzky_Golay Filter平滑后去噪，去噪的情况包括但不限于修正Savitzky_Golay Filter平滑后负数的值等。计算平滑后曲线的波峰顶点与宽度，可以用scipy.signal.find_peaks等类似方法；在进行平滑处理之后的曲线中包括非植被像素时，对直线集合进行偏移，以使得目标曲线中不包含非植被像素，具体地，必要的情况下，可以对直线集合在一定邻域范围内左右偏移，偏移的触发条件可以是邻域内原始累加曲线存在累加值为零的横轴点。偏移的原因是有可能一个峰值点经过平滑之后，距离不是原来的、真正的零值点有所偏移，那它的平移是有限的，设定一个相邻距离阈值(邻域)，在峰值的邻域内搜索平滑前累加曲线的累加值为零的点，此外，如果最左最右缺少边界线，可以补加边界线。
本申请的一些实施例中，从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像之后，可以从第二二值化图像中提取样本数据，其中，该样本数据中包括：作物连通域；对样本数据中的作物连通域进行标记，得到作物标签；基于作物连通域和为作物连通域分配的标签对作物识别模型进行训练，得到训练后的作物识别模型，上述作物识别模型可以基于作物连通域进行识别，进而实现识别缺失作物的目的。
需要说明的是，可以采用人工方式对作物进行标记，也可以采用深度学习算法进行标记进而生成作物标签，其中作物标签可以是不规则图形连通域、高斯圆、矩形框集合，也可以是多边形等符合网络的输入的形状。
缺苗标记可以为在检测出缺苗的两个相邻连通域，或两个相邻连通域之间进行的标记。在一个具体的示例中，可以对检测出缺苗的两个相邻连通域连接线段，作为缺苗标记。在另一个具体的例子中，可以在检测出缺苗的两个相邻连通域之间的缺苗区间内，按照理想相邻距离补充任一其他位置作物连通域，理想相邻距离可以根据各种植行区域内相邻作物连通域之间的距离的平均值或众数获得。本申请实施例对进行缺苗标记的方式和具体过程不进行限制。
在本申请的一些实施例中，根据各种植行区域内相邻作物连通域之间的距离进行缺苗标记。这样设置的好处在于，按照相邻距离判断是否缺苗，对于并非严格进行直线种植的作物，同样可以进行缺苗标记，适用范围更加广泛。
本申请一种可选的实施例中，从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像，可以先确定区域图像中各个像素点的绿度指数；对于每个像素点， 比较绿度指数和绿度阈值的大小；依据比较结果确定像素点是否为植被区域中的像素点；统计属于植被区域中的像素点，并基于统计结果确定第一二值化图像，具体地，以超绿指数(Excess Green)为例，ExG＝2*Green–Red–Blue，其中公式右边是G、R、B三通道的某像素值，计算超绿指数后，通过阈值分离出植被区域，像素点的超绿指数大于或等于阈值的为非零值，像素点的超绿指数小于阈值的置零，其中大于或等于阈值的非零值的灰度值设为255，小于阈值的置零值灰度值设为0，生成第一二值化图，
需要说明的是，上述公式ExG＝2*Green–Red–Blue在实际使用需要进行归一化等数学处理，阈值可以根据作物的种植面积、类型、形状大小、种植经验等人为设定，也可以通过Ostu法获得。
本申请的一些实施例中，从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像，具体他，获取第一二值化图像中各个植被连通域的面积；对于每个植被连通域，确定植被连通域的面积是否属于预设取值范围，如果属于预设取值范围，则确定植被连通域为作物连通域；基于确定的作物连通域生成二值化图像。
本申请的一些的施例中，筛选植被连通域的方法，包括但不限于以下方法：人为设定阈值法、统计筛选法、形状筛选法、纹理筛选法等。
具体地，例如，通过形状筛选法筛选植被连通域，例如，田边停放了一辆绿颜色的卡车，作物为形状近似为圆形的苹果树，便可以通过几何分析后，确定这辆车不属于植被连通域。
具体地，例如，通过统计筛选法通过统计所有的连通域面积，通过众数、密集区间等方法确定作物面积范围，在面积范围内的植被连通域留下筛选，例如，植被连通域面积区间为[20，50](单位/㎡)，则符合这个面积区间的植被连通域被留下，例如，植被连通域面积的大小众数为30㎡，则可以将面积为30平方米的连通域留下。
需要说明的是，为了提高筛选的准确性，上述筛选方式可以结合起来使用，例如，在田边停放的一辆车，符合植被连通域的超绿指数，形状也与作物的形状相似，则可以进一步结合面积区间，来判断上述车辆是否为植被连通域。
本申请一种可选的实施例中，确定种植行区域处于作物缺失状态之前，可以统计目标区域中任意两个相邻的作物连通区域之间的距离集合；计算距离集合中的所有距离的平均值，并基于平均值确定预设阈值，其中，所述距离为所述相邻的作物连通域质心之间的距离；也可以直接接受用户输入的预设阈值。
本申请的一些实施例中，基于上述平均值确定预设阈值，可以将平均值确定为预设阈值；确定预设阈值的方法也可以为：统计目标区域中各个作物连通区域的半径，得到作物连通域的平均半径，基于平均半径和上述平均值确定预设阈值，具体地，可以为平均半径与平均值求和后得到的结果作为预设阈值。
本申请一种可选的实施例中，确定种植行区域处于作物缺失状态之后，可以确定种植行区域中处于作物缺失状态的位置；并在处于作物缺失状态的位置处生成用于指示作物缺失状态的标记。
本本申请一种可选的实施例中，在处于作物缺失状态的位置处生成用于指示作物缺失状态的标记，包括：以处于缺失状态的位置为中心，以作物连通域的平均半径为半径，生成圆形标记，并将该圆形标记作为用于指示作物缺失状态的标记；或者以处于缺失状态的位置为中心，以作物连通域的平均半径的N倍为边长生成矩形标记，并将该矩形标记作为用于指示作物缺失状态的标记，为了防止生成的矩形标记与相邻的作物连通域的边界产生重合或者覆盖相邻的作物连通域，其中，0＜N≤2。
在本申请的一些实施例中，可以通过以下方式确定端部作物连通域与所在种植行区域的区域边界之间的距离大于预设阈值：获取与所述第二二值化图像匹配的作物边界；如果确定端部作物连通域处于所在种植行区域的极限位置，则获取与所述端部作物连通域匹配的端部作物边界；沿所述种植行主方向，计算所述端部作物连通域的质心与所述端部作物边界之间的距离值；如果距离超过所述缺苗阈值距离，则确定端部作物连通域与所在种植行区域的区域边界之间的距离大于所述预设阈值。
其中，极限位置可以指种植行区域的两端，端部作物边界可以指距离端部作物连通域距离最近的作物边界。在本申请的一些实施例中，当作物连通域位于种植行区域两端处，需要判断该作物连通域与距离其最近的作物边界之间的距离，是否大于缺苗阈值距离。
本申请的一些实施例中，获取目标区域的区域图像，可以是接收来自无人机拍摄的目标区域的区域图像。
图5是根据本申请实施例的一种目标区域中作物缺失的检测装置，如图5所示，该装置包括：
获取模块50，设置为获取目标区域的区域图像；
第一生成模块52，设置为从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；
第二生成模块54，设置为从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；
第一确定模块56，设置为对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距，其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；
第二确定模块58，设置为根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
上述目标区域中作物缺失的检测装置中，包括：获取模块50，设置为获取目标区域的区域图像，第一生成模块52，设置为从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；第二生成模块54，设置为从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；第一确定模块56，设置为对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的距离；第二确定模块58，设置为在距离大于预设阈值时，确定种植行区域处于作物缺失状态，达到了识别缺失作物的目的，从而实现了及时、高效，准确 地识别缺失作物的技术效果，进而解决了人工检测缺失作物，需要大量人力参与、耗时费力，且容易出现人为失误的技术问题。
需要说明的是，小于或者等于预设阈值时，确定种植行区域不处于作物缺失状态。
图6是根据本申请实施例的一种无人机，如图6所示，该无人机包括：图像采集装置60，设置为获取目标区域的区域图像；处理器62，设置为从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距，其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
需要说明的是，小于或者等于预设阈值时，确定种植行区域不处于作物缺失状态。
图7是根据本申请实施例的一种缺失作物数的计算方法，该方法包括如下步骤：
步骤S702：获取目标区域的区域图像；
步骤S704，从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；
步骤S706，从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；
步骤S708，对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的距离；
步骤S710，在距离大于预设阈值时，确定种植行区域处于作物缺失状态，生成作物缺失标记；
步骤S712，依据缺失标记，识别缺苗位置，计算缺失的作物数目。
上述缺失作物数的计算方法中，首先，获取目标区域的区域图像，其次，从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像，再从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像，再对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的距离，再在距离大于预设阈值时，确定种植行区域处于作物缺失状态，生成作物缺失标记，最后，依据缺失标记，识别缺苗位置，计算缺失的作物数目。
需要说明的是，小于或者等于预设阈值时，确定种植行区域不处于作物缺失状态
图8是根据本申请实施例的一种缺失作物数的计算装置，该装置包括：获取模块80，设置为获取目标区域的区域图像；识别模块82，设置为从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像对于所述第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距；其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；计算模块84，设置为根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
需要说明的是，小于或者等于预设阈值时，确定种植行区域不处于作物缺失状态。
图9是根据本申请实施例的另一种作物缺失的检测方法，如图9所示，该方法包括如下步骤：
步骤S902，获取目标区域的区域图像；上述区域图像包括但不限于农田图像，例如可以为对农田的俯拍图。
步骤S904，将区域图像输入至作物识别模型进行分析，得到目标区域中的各个作物连通域，并基于作物连通域生成二值化图像，其中，作物识别模型为基于多组数据训练得到的，多组数据中的每组数据中均包括样本图像，以及用于标记样本图像中的连通域的标签；
步骤S906，确定二值化图像中每个种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物联通域与种植行区域的便捷之间的临界间距，其中，端部作物联通域在种植行区域中仅有一个作物联通域相邻；
步骤S908，根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
上述作物缺失的检测方法中，首先，获取目标区域的区域图像；其次，将区域图像输入至作物识别模型进行分析，得到目标区域中的各个作物连通域，并基于作物连通域生成二值化图像，其中，作物识别模型为基于多组数据训练得到的，多组数据中的每组数据中均包括样本图像，以及用于标记样本图像中的连通域的标签；最后，确定每个种植行区域中任意两个相邻的作物连通域之间的距离，在距离大于预设阈值时，确定种植行区域处于作物缺失状态，从而实现了及时、高效，准确地识别缺失作物的技术效果，进而解决了人工检测缺失作物，需要大量人力参与、耗时费力，且容易出现人为失误的技术问题。
需要说明的是，小于或者等于预设阈值时，确定种植行区域不处于作物缺失状态。
本申请一种可选的实施例中，确定每个种植行区域中任意两个相邻的作物连通域之间的距离，包括：依据预设阈值对种植行区域进行双向检测，其中，该双向检测为依据一个方向对种植行区域中的各个作物连通域之间的距离与预设阈值进行比较后，再按照方向的反方向对各个作物连通域之间的距离与预设阈值进行比较；距离大于预设阈值时，确定种植行区域处于作物缺失状态，包括：对于任意两个相邻的作物连通域，在双向检测结果中任意一个方向的检测结果指示任意两个相邻的作物连通域之间的距离大于预设阈值时，确定种植行区域处于作物缺失状态。
本申请一种可选的实施例中，将区域图像输入至作物识别模型进行分析，得到目标区域中的各个作物连通域之后，方法还包括：确定目标区域中作物种植行的主方向；基于目标区域中各个种植行在主方向的长度生成二值化图像的外接矩形；并以矩形的宽度为纵坐标，长度方向为横坐标，建立坐标系；确定主方向上非植被像素的个数，并基于非植被像素的个数，在坐标系中生成用于指示非植被像素的个数分布情况的曲线；依据曲线的波峰顶点、主方向对应的直线集合中的直线，切割第二二值化图像，得到种植行区域，其中，直线集合中的直线用于指示相邻种植行之间，非植被像素所在的位置点集合。
图10是本申请一种可选的作物缺失的检测原理示意图，如图10所示，该流程主 要包括以下步骤：
首先，俯拍农田的大量照片，形成农田照片集，利用算法提取照片中植被所在所在区域，并对所有植被连通域进行筛选，留下作物的连通域，生成作物的二值化图，进而生成作物标签，或者直接通过人工标注的方法，生成作物标签，得到农田照片与生成作物标签成对存储的训练数据集、测试数据集、验证数据集；其次，通过深度度网络对训练数据集进行训练，得到深度学习模型，验证集用于调超参数，选出效果最优模型所对应的参数，测试集用来衡量最优的性能；最终得到作物分割模型。
作物分割模型用于识别待执行图片，得到作物的分割图，利用算法得到作物种植行的的主方向，其中上述算法包括但不限于利用霍夫检测直线的方法，具体地，首先将作物分割图对应的二值化图转换到极坐标系霍夫空间，通过霍夫空间累加器返回靠前的直线的极坐标角度θ、半径ρ列表的结果，对列表数据进行统计分析，获得种植行的主方向。统计分析方法包括但不限于选取出现次数最多的角度作为种植行的主方向，返回结果较少时选取排名第一的角度作为种植行的主方向，然后在以主方向为高度生成二值化图的外接矩形，以外接矩形宽度方向(主方向)为纵坐标，长度方向为横坐标，累加主方向上非植被像素个数，得到累加曲线，因为实际上种植行的方向(主方向)不一定与图片观察角度垂直，外接矩形的目的是沿着主方向累加零像素，以得到累加曲线进行波形分析，对累加曲线进行平滑处理后，计算平滑后曲线的波峰顶点，根据波峰顶点、主方向确定的直线集合，切割出二值化图种植行区域，逐行对所有作物连通域进行相邻距离统计，得出作物缺失半径；或者人为设定作物缺失半径。这里必要统计的是连通域两个质心的平均相邻距离，进一步可以加上统计连通域的平均半径，如此一来，相邻连通域的出现范围应该在[相邻质心距离-平均半径，相邻质心距离+平均半径]；逐个种植行区域内逐个连通域以缺苗距离为半径双向搜索是否存在连通域，若不存在，判断为缺失，并生成作物缺失标记。缺失标记的生成，可以是质心在当前连通域质心的行方向直线上的距离为平均相邻距离的点上，半径为平均半径的圆，或者边长为N倍半径的矩形，其中，0＜N≤2。
在作物的耕种过程中，由于农机故障等人为因素，一般都会导致某些区域内存在缺失作物苗的现象，因此，根据本申请实施例的另一方面，还提供了一种补种方法，具体地，可以采用任意一种作物缺失的检测方法确定缺苗区域；然后，根据缺苗区域生成农机作业的推荐路径，对缺苗区域进行补种，例如，当农机作业时，可以依据作物缺失的检测方法确定出缺苗区域，然后农机就可以直接这些若干个缺苗区域，进行补种作业，而不必对整个作物区域进行一一排查再进行补种。
在本申请的一些实施例中，缺苗区域的确定过程包括：基于权利要求1至20任一项所述的作物缺失的检测方法，对目标区域的区域图像进行处理，确定所述目标区域所包含的种植行区域是否处于作物缺失状态；确定种植行区域处于作物缺失状态时，生成作物缺失标记；根据作物缺失标记在区域图像二值图中的图像位置，以及与区域图像二值图匹配的地理位置信息，确定各作物缺失标记对应的地理位置信息，以得到缺苗区域。
根据本申请实施例的另一方面，还提供了一种作业路线规划方法，具体地，采用任意一种作物缺失的检测方法确定缺苗区域；根据缺苗区域确定作业路线，例如，当农机在进行作业时，可以首先识别出依据作物缺失的检测方法确定出缺苗区域，然后 农机就可以依据这些若干个缺苗区域，规划出避开缺苗区域的路径。
为了方便各种类型的设备进行作业，节省各种设备作业的时间和成本，根据本申请实施例的另一方面，还提供了一种作业方法，具体地，作业设备首先执行任意一种作物缺失的检测方法确定缺苗区域；然后作业设备标记缺苗区域，确定缺苗位置；最后，当作业设备所在位置位于缺苗位置时，则控制设备停止在缺苗位置进行作业。
需要说明的是，上述设备包括但不限于以下类型：喷洒设备、播撒设备和采收设备。
具体地，例如喷洒设备在喷洒农药时，可以依据作物缺失的检测方法得到缺苗区域，当喷洒设备落入缺苗区域的范围时，停止喷洒作业；又例如，当播撒设备在播撒固态化肥时，当播撒设备落入缺苗区域的范围时，停止播撒作业。再例如，当采收设备收割作物时，行驶到缺苗区域时，可以自行收起收割装置，暂停收割作业。
为了更加准确的估算某块区域内作物的产量，根据本申请实施例的另一方面，还提供了一种产量测算方法，具体地：首先，可采用上述作物缺失的检测方法确定缺苗区域；然后，根据缺苗区域确定非缺苗区域的面积；最后，通过非缺苗区域的单位面积产量与非缺苗区域的总面积的乘积确定作物区域总产量，例如，某块农田的总面积为1公顷、非缺苗区域单位面积的产量为1吨，经过上述作物缺失的检测方法识别出了这片农田中若干缺苗区域，然后再得到缺苗区域面积大小为0.1公顷，此时，非缺苗区域的面积大小为农田的总面积减去缺苗区域面积等于0.9公顷，然后再乘以上述单位产量1吨后，可以得到非缺苗区域的作物产量为0.9吨，即这块农田的产量为0.9吨。
根据本申请实施例的另一方面，还提供了一种非易失性存储介质，非易失性存储介质包括存储的程序，其中，在程序运行时控制非易失性存储介质所在设备执行任意一种作物缺失的检测方法，或任意一种缺失作物数的计算方法，或任意一种作业路径规划方法，或任意一种补种方法，或任意一种作业方法，或任意一种产量测算方法。
在本申请的一些实施例中，上述非易失性存储介质用于存储执行以下功能的程序指令，实现以下功能：
获取目标区域的区域图像；从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距，其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
根据本申请实施例的另一方面，还提供了一种处理器，其中，处理器用于运行程序，其中，程序运行时执行任意一种作物缺失的检测方法，或任意一种缺失作物数的计算方法，或任意一种作业路径规划方法，或任意一种补种方法，或任意一种作业方法，或任意一种产量测算方法。
在本申请的一些实施例中，上述处理器用于调用存储器中的程序指令，实现以下 功能：
获取目标区域的区域图像；从区域图像中提取植被区域，并生成植被区域对应的第一二值化图像；从第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；对于第二二值化图像中的每个种植行区域，确定种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或种植行区域中端部作物连通域与种植行区域的边界之间的临界间距，其中，端部作物连通域在种植行区域中仅与一个作物连通域相邻；根据作物间距或临界间距确定种植行区域是否处于作物缺失状态。
上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。
在本申请的上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述的部分，可以参见其他实施例的相关描述。
在本申请所提供的几个实施例中，应该理解到，所揭露的技术内容，可通过其它的方式实现。其中，以上所描述的装置实施例仅仅是示意性的，例如所述单元的划分，可以为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，单元或模块的间接耦合或通信连接，可以是电性或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可为个人计算机、服务器或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、移动硬盘、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述仅是本申请的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本申请原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本申请的保护范围。

Claims (32)

  1. 一种作物缺失的检测方法，包括：
    获取目标区域的区域图像；
    从所述区域图像中提取植被区域，并生成所述植被区域对应的第一二值化图像；
    从所述第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；
    对于所述第二二值化图像中的每个种植行区域，确定所述种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或所述种植行区域中端部作物连通域与所述种植行区域的边界之间的临界间距；其中，所述端部作物连通域在所述种植行区域中仅与一个作物连通域相邻；
    根据所述作物间距或所述临界间距确定所述种植行区域是否处于作物缺失状态。
  2. 根据权利要求1所述的方法，其中，根据所述作物间距确定所述种植行区域是否处于作物缺失状态，包括：
    依据预设阈值对所述种植行区域进行双向检测，其中，该双向检测为依据一个方向对所述种植行区域中的各个作物连通域之间的距离与所述预设阈值进行比较后，再按照所述方向的反方向对各个作物连通域之间的距离与所述预设阈值进行比较；
    对于所述任意两个相邻的作物连通域，在双向检测结果中任意一个方向的检测结果指示所述任意两个相邻的作物连通域之间的作物间距大于所述预设阈值时，确定所述种植行区域处于作物缺失状态。
  3. 根据权利要求1或2所述的方法，其中，根据所述临界间距确定所述种植行区域是否处于作物缺失状态，包括：
    获取与所述第二二值化图像匹配的作物边界；
    从所述作物边界中获取与所述端部作物连通域匹配的端部作物边界；
    沿所述种植行区域的种植行主方向，计算所述端部作物连通域的质心与所述端部作物边界之间的临界间距；
    在所述临界间距大于预设阈值时，确定所述种植行区域处于作物缺失状态。
  4. 根据权利要求1所述的方法，其中，所述第二二值化图像中的每个种植行区域的 确定过程包括：
    确定所述目标区域中作物种植行的主方向；
    基于所述目标区域中各个种植行在主方向的长度生成所述第二二值化图像的外接矩形；并以所述矩形的宽度为纵坐标，长度方向为横坐标，建立坐标系；
    确定所述主方向上非植被像素的个数，并基于所述非植被像素的个数，在所述坐标系中生成用于指示非植被像素的个数分布情况的曲线；
    依据所述曲线的波峰顶点、所述主方向对应的直线集合中的直线，切割第二二值化图像，得到所述种植行区域，其中，所述直线集合中的直线用于指示相邻种植行之间，所述非植被像素所在的位置点集合。
  5. 根据权利要求4所述的方法，其中，确定所述目标区域中作物种植行的主方向，包括：
    对所述第二二值化图像进行霍夫变换，并通过霍夫空间累加器输出所述第二二值化图像在霍夫空间中的极坐标系参数对列表，基于所述极坐标系参数对列表中的角度参数确定所述作物种植行的主方向，其中，所述极坐标系参数对列表中的包括至少一个极坐标系参数对，每个极坐标系参数对中均包括角度和半径。
  6. 根据权利要求5所述的方法，其中，通过霍夫空间累加器输出所述第二二值化图像在霍夫空间中的极坐标系参数对列表，基于所述极坐标系参数对列表中的角度参数确定所述作物种植行的主方向，包括：
    在所述极坐标系参数对列表中的极坐标参数对的数量大于预定值时，将所述极坐标参数对列表中出现次数最多的角度作为所述作物种植行的主方向；或者，
    在所述极坐标系参数对列表中的极坐标参数对的数量小于预定值时，将所述极坐标系参数对列表中位于首位的极坐标角度作为所述作物种植行的主方向。
  7. 根据权利要求4所述的方法，其中，依据所述曲线的波峰顶点、所述主方向对应的直线集合中的直线，切割第二二值化图像，得到所述种植行区域之前，所述方法还包括：
    对所述曲线进行平滑处理，得到所述平滑处理后的目标曲线；
    在进行平滑处理之后的所述目标曲线中包括非植被像素时，对所述直线集合进行偏移，以使得所述目标曲线中不包含所述非植被像素。
  8. 根据权利要求4或7所述的方法，其中，依据所述曲线的波峰顶点、所述主方向 对应的直线集合中的直线，切割第二二值化图像，得到所述种植行区域，包括：
    获取所述曲线中的各波峰顶点的横坐标，沿所述种植行主方向生成直线；
    在所述第二二值图中，将相邻两条直线之间的区域作为所述种植行区域。
  9. 根据权利要求8所述的方法，其中，在获取所述曲线中的各波峰顶点的横坐标之后，所述方法还包括：
    依次获取一个目标波峰顶点的横坐标作为当前处理横坐标，并获取所述当前处理横坐标关联的邻域横坐标集合；
    如果所述邻域横坐标集合中目标邻域横坐标满足临界点条件，则将所述目标波峰顶点的横坐标替换为所述目标邻域横坐标；
    返回执行依次获取一个目标波峰顶点的横坐标作为当前处理横坐标的操作，直至完成对全部波峰顶点的横坐标的处理。
  10. 根据权利要求1所述的方法，其中，从所述第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像之后，所述方法还包括：
    从所述第二二值化图像中提取样本数据，其中，该样本数据中包括：作物连通域；
    对所述样本数据中的作物连通域进行标记，得到作物标签；
    基于所述作物连通域和为所述作物连通域分配的标签对作物识别模型进行训练，得到训练后的作物识别模型。
  11. 根据权利要求1所述的方法，其中，从所述区域图像中提取植被区域，并生成所述植被区域对应的第一二值化图像，包括：
    确定所述区域图像中各个像素点的绿度指数；
    对于每个像素点，比较所述绿度指数和绿度阈值的大小；
    依据比较结果确定所述像素点是否为所述植被区域中的像素点；
    统计属于所述植被区域中的像素点，并基于统计结果确定所述第一二值化图像。
  12. 根据权利要求1所述的方法，其中，从所述第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像，包括；
    获取所述第一二值化图像中各个植被连通域的面积；
    对于每个植被连通域，确定所述植被连通域的面积是否属于预设取值范围，如果属于所述预设取值范围，则确定所述植被连通域为作物连通域；
    基于确定的作物连通域生成所述第二二值化图像。
  13. 根据权利要求2所述的方法，其中，所述预设阈值通过以下任一种方式获取：
    第一种方式：统计所述目标区域中任意两个相邻的作物连通区域质心之间的距离集合；计算所述距离集合中的所有距离的平均值，并基于所述平均值确定所述预设阈值；
    第二种方式：接收用户输入的所述预设阈值。
  14. 根据权利要求13所述的方法，其中，基于所述平均值确定所述预设阈值，包括：
    将所述平均值确定为所述预设阈值；
    或者，统计所述目标区域中各个作物连通区域的半径，得到所述作物连通域的平均半径；基于所述平均半径和所述平均值确定所述预设阈值。
  15. 根据权利要求1所述的方法，其中，确定所述种植行区域处于作物缺失状态之后，所述方法还包括：
    确定所述种植行区域中处于作物缺失状态的位置；并在处于作物缺失状态的位置处生成用于指示作物缺失状态的标记。
  16. 根据权利要求15所述的方法，其中，在处于作物缺失状态的位置处生成用于指示作物缺失状态的标记，包括：
    以所述处于缺失状态的位置为中心，以所述作物连通域的平均半径为半径，生成圆形标记，并将该圆形标记作为所述用于指示作物缺失状态的标记；或者
    以所述处于缺失状态的位置为中心，以所述作物连通域的平均半径的N倍为边长生成矩形标记，并将该矩形标记作为所述用于指示作物缺失状态的标记，其中，0＜N≤2。
  17. 根据权利要求1至16中任意一项所述的方法，其中，获取目标区域的区域图像，包括：
    接收无人机拍摄的所述目标区域的区域图像。
  18. 一种作物缺失的检测方法，包括：
    获取目标区域的区域图像；
    将所述区域图像输入至作物识别模型进行分析，得到所述目标区域中的各个作物连通域，并基于所述作物连通域生成二值化图像，其中，所述作物识别模型为基于多组数据训练得到的，多组数据中的每组数据中均包括样本图像，以及用于标记样本图像中的连通域的标签；
    确定二值化图像中每个种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或所述种植行区域中端部作物连通域与所述种植行区域的边界之间的临界间距；其中，所述端部作物连通域在所述种植行区域中仅与一个作物连通域相邻；
    根据所述作物间距或所述临界间距确定所述种植行区域是否处于作物缺失状态。
  19. 根据权利要求18所述的方法，其中，根据所述作物间距确定所述种植行区域是否处于作物缺失状态，包括：
    依据预设阈值对所述种植行区域进行双向检测，其中，该双向检测为依据一个方向对所述种植行区域中的各个作物连通域之间的距离与所述预设阈值进行比较后，再按照所述方向的反方向对各个作物连通域之间的距离与所述预设阈值进行比较；
    对于所述任意两个相邻的作物连通域，在双向检测结果中任意一个方向的检测结果指示所述任意两个相邻的作物连通域之间的作物间距大于所述预设阈值时，确定所述种植行区域处于作物缺失状态。
  20. 根据权利要求18所述的方法，其中，所述二值化图像中的每个种植行区域的确定过程包括：
    确定所述目标区域中作物种植行的主方向；
    基于所述目标区域中各个种植行在主方向的长度生成所述二值化图像的外接矩形；并以所述矩形的宽度为纵坐标，长度方向为横坐标，建立坐标系；
    确定所述主方向上非植被像素的个数，并基于所述非植被像素的个数，在所述坐标系中生成用于指示非植被像素的个数分布情况的曲线；
    依据所述曲线的波峰顶点、所述主方向对应的直线集合中的直线，切割第二二值化图像，得到所述种植行区域，其中，所述直线集合中的直线用于指示相邻种植行之间，所述非植被像素所在的位置点集合。
  21. 一种缺失作物数的计算方法，包括：
    基于权利要求1至20任一项所述的作物缺失的检测方法，对目标区域的区域图像进行处理，确定所述目标区域所包含的种植行区域是否处于作物缺失状态；
    确定所述种植行区域处于作物缺失状态时，生成作物缺失标记；
    依据所述作物缺失标记，计算缺失的作物数目。
  22. 一种作业路线规划方法，包括：
    基于权利要求1至20任一项所述的作物缺失的检测方法，确定缺苗区域；
    根据所述缺苗区域生成作业路线，其中，所述作业路线经过所述缺苗区域。
  23. 一种补种方法，包括：
    基于权利要求1至20任一项所述的作物缺失的检测方法，确定缺苗区域；
    根据所述缺苗区域生成作业路线；
    向作业设备发送所述作业路线，其中，所述作业路线用于指示作业设备对所述缺苗区域进行作物补种。
  24. 根据权利要求23所述的补种方法，其中，缺苗区域的确定过程包括：
    基于权利要求1至20任一项所述的作物缺失的检测方法，对目标区域的区域图像进行处理，确定所述目标区域所包含的种植行区域是否处于作物缺失状态；
    确定所述种植行区域处于作物缺失状态时，生成作物缺失标记；
    根据作物缺失标记在区域图像二值图中的图像位置，以及与区域图像二值图匹配的地理位置信息，确定各作物缺失标记对应的地理位置信息，以得到缺苗区域。
  25. 根据权利要求24所述的补种方法，其中，根据所述缺苗区域生成补种作业路线，包括：
    根据各所述作物缺苗标记对应的地理位置信息，生成与所述种植区域匹配的补种作业路线。
  26. 一种作业方法，应用于作业设备，所述作业设备包括以下至少之一：喷洒设备、播撒设备和采收设备；所述方法包括：
    确定作业设备的当前位置是否位于目标区域的缺苗位置，如果所述作业设备位于所述缺苗位置，则停止在所述缺苗位置进行作业，其中，所述缺苗位置基于权利要求1至20任一项所述的作物缺失的检测方法确定得到。
  27. 一种产量测算方法，包括：
    基于权利要求1至20任一项所述的作物缺失的检测方法确定目标区域中的缺苗区域；
    根据所述缺苗区域确定所述目标区域中非缺苗区域的面积；
    基于所述非缺苗区域的单位面积产量与所述非缺苗区域的总面积确定所述目标区域的总产量。
  28. 一种目标区域中作物缺失的检测装置，包括：
    获取模块，设置为获取目标区域的区域图像；
    第一生成模块，设置为从所述区域图像中提取植被区域，并生成所述植被区域对应的第一二值化图像；
    第二生成模块，设置为从所述第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；
    第一确定模块，设置为对于所述第二二值化图像中的每个种植行区域，确定所述种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或所述种植行区域中端部作物连通域与所述种植行区域的边界之间的临界间距；其中，所述端部作物连通域在所述种植行区域中仅与一个作物连通域相邻；
    第二确定模块，设置为根据所述作物间距或所述临界间距确定所述种植行区域是否处于作物缺失状态。
  29. 一种无人机，包括：
    图像采集装置，设置为获取目标区域的区域图像；
    处理器，设置为从所述区域图像中提取植被区域，并生成所述植被区域对应的第一二值化图像；从所述第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；对于所述第二二值化图像中的每个种植行区域，确定所述种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或所述种植行区域中端部作物连通域与所述种植行区域的边界之间的临界间距；其中，所述端部作物连通域在所述种植行区域中仅与一个作物连通域相邻；根据所述作物间距或所述临界间距确定所述种植行区域是否处于作物缺失状态。
  30. 一种缺失作物数的计算装置，其中，包括
    获取模块，设置为获取目标区域的区域图像；
    识别模块，设置为从所述区域图像中提取植被区域，并生成所述植被区域对应的第一二值化图像；从所述第一二值化图像的植被连通域中确定作物连通域，并基于该作物连通域生成第二二值化图像；对于所述第二二值化图像中的每个种植行区域，确定所述种植行区域中任意两个相邻的作物连通域之间的作物间距，和/或所述种植行区域中端部作物连通域与所述种植行区域的边界之间的临界间距；其中，所述端部作物连通域在所述种植行区域中仅与一个作物连通域相邻；
    计算模块，设置为根据所述作物间距或所述临界间距确定所述种植行区域是否处于作物缺失状态。
  31. 一种非易失性存储介质，所述非易失性存储介质包括存储的程序，其中，在所述程序运行时控制所述非易失性存储介质所在设备执行权利要求1至17中任意一项所述的作物缺失的检测方法，或权利要求18至20中任一项所述的作物缺失的检测方法，或权利要求21所述的缺失作物数的计算方法，或权利要求22所述的作业路线规划方法，或权利要求23至25中任一项所述的补种方法，或权利要求26所述的作业方法，或权利要求27所述的产量测算方法。
  32. 一种处理器，所述处理器用于运行程序，其中，所述程序运行时执行权利要求1至18中任意一项所述的作物缺失的检测方法，或权利要求18至20中任一项所述的作物缺失的检测方法，或权利要求21所述的缺失作物数的计算方法，或权利要求22所述的作业路线规划方法，或权利要求23至25中任一项所述的补种方法，或权利要求26所述的作业方法，或权利要求27所述的产量测算方法。
PCT/CN2021/099863 2020-06-12 2021-06-11 作物缺失的检测方法及检测装置 WO2021249560A1 (zh)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN202010538238.5 2020-06-12
CN202010537090.3A CN113807128B (zh) 2020-06-12 2020-06-12 一种缺苗标记方法、装置、计算机设备和存储介质
CN202010538238.5A CN113807138A (zh) 2020-06-12 2020-06-12 作物缺失的检测方法及检测装置
CN202010537090.3 2020-06-12

Publications (1)

Publication Number Publication Date
WO2021249560A1 true WO2021249560A1 (zh) 2021-12-16

Family

ID=78846884

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/099863 WO2021249560A1 (zh) 2020-06-12 2021-06-11 作物缺失的检测方法及检测装置

Country Status (1)

Country Link
WO (1) WO2021249560A1 (zh)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114359260A (zh) * 2022-01-18 2022-04-15 成都理工大学 一种烟条表面的缺陷检测方法及装置
CN114612376A (zh) * 2022-01-18 2022-06-10 华南理工大学 自动图像分辨率测量方法、装置、计算机设备和存储介质
CN114913180A (zh) * 2022-07-19 2022-08-16 海门市芳华纺织有限公司 一种棉布筘痕缺陷的智能检测方法
CN115272298A (zh) * 2022-09-19 2022-11-01 江苏网进科技股份有限公司 基于道路监控的城市路面维护监察方法及系统
CN115358998A (zh) * 2022-08-22 2022-11-18 法博思(宁波)半导体设备有限公司 一种随机阵列图片中点坐标的获取方法及系统
CN116453003A (zh) * 2023-06-14 2023-07-18 之江实验室 一种基于无人机监测智能识别水稻生长势的方法和系统
CN117218549A (zh) * 2023-11-07 2023-12-12 山东力加力钢结构有限公司 一种公路桥梁的状态评估方法及系统
CN117635719A (zh) * 2024-01-26 2024-03-01 浙江托普云农科技股份有限公司 基于多传感器融合的除草机器人定位方法、系统及装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106250831A (zh) * 2016-07-22 2016-12-21 北京小米移动软件有限公司 图像检测方法、装置和用于图像检测的装置
CN109658405A (zh) * 2018-12-20 2019-04-19 中国气象局气象探测中心 一种农作物实景观测中图像数据质量控制方法及系统
CN109886094A (zh) * 2019-01-08 2019-06-14 中国农业大学 一种作物苗情苗势采集分析方法和装置
US20190258859A1 (en) * 2016-09-07 2019-08-22 Precision Hawk Usa, Inc. Systems and methods for mapping emerged plants
CN110765927A (zh) * 2019-10-21 2020-02-07 广西科技大学 一种植被群落中伴生杂草的识别方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106250831A (zh) * 2016-07-22 2016-12-21 北京小米移动软件有限公司 图像检测方法、装置和用于图像检测的装置
US20190258859A1 (en) * 2016-09-07 2019-08-22 Precision Hawk Usa, Inc. Systems and methods for mapping emerged plants
CN109658405A (zh) * 2018-12-20 2019-04-19 中国气象局气象探测中心 一种农作物实景观测中图像数据质量控制方法及系统
CN109886094A (zh) * 2019-01-08 2019-06-14 中国农业大学 一种作物苗情苗势采集分析方法和装置
CN110765927A (zh) * 2019-10-21 2020-02-07 广西科技大学 一种植被群落中伴生杂草的识别方法

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114359260B (zh) * 2022-01-18 2023-11-03 成都理工大学 一种烟条表面的缺陷检测方法及装置
CN114612376A (zh) * 2022-01-18 2022-06-10 华南理工大学 自动图像分辨率测量方法、装置、计算机设备和存储介质
CN114359260A (zh) * 2022-01-18 2022-04-15 成都理工大学 一种烟条表面的缺陷检测方法及装置
CN114612376B (zh) * 2022-01-18 2024-04-05 华南理工大学 基于双丝型像质计的一键式图像测量方法、装置及存储介质
CN114913180A (zh) * 2022-07-19 2022-08-16 海门市芳华纺织有限公司 一种棉布筘痕缺陷的智能检测方法
CN115358998A (zh) * 2022-08-22 2022-11-18 法博思(宁波)半导体设备有限公司 一种随机阵列图片中点坐标的获取方法及系统
CN115272298A (zh) * 2022-09-19 2022-11-01 江苏网进科技股份有限公司 基于道路监控的城市路面维护监察方法及系统
CN115272298B (zh) * 2022-09-19 2022-12-13 江苏网进科技股份有限公司 基于道路监控的城市路面维护监察方法及系统
CN116453003A (zh) * 2023-06-14 2023-07-18 之江实验室 一种基于无人机监测智能识别水稻生长势的方法和系统
CN116453003B (zh) * 2023-06-14 2023-09-01 之江实验室 一种基于无人机监测智能识别水稻生长势的方法和系统
CN117218549A (zh) * 2023-11-07 2023-12-12 山东力加力钢结构有限公司 一种公路桥梁的状态评估方法及系统
CN117218549B (zh) * 2023-11-07 2024-03-05 山东力加力钢结构有限公司 一种公路桥梁的状态评估方法及系统
CN117635719A (zh) * 2024-01-26 2024-03-01 浙江托普云农科技股份有限公司 基于多传感器融合的除草机器人定位方法、系统及装置
CN117635719B (zh) * 2024-01-26 2024-04-16 浙江托普云农科技股份有限公司 基于多传感器融合的除草机器人定位方法、系统及装置

Similar Documents

Publication Publication Date Title
WO2021249560A1 (zh) 作物缺失的检测方法及检测装置
Malambo et al. Automated detection and measurement of individual sorghum panicles using density-based clustering of terrestrial lidar data
CN102704215B (zh) 基于dst文件解析与机器视觉结合的绣布自动切割方法
US10127450B2 (en) Plot gap identification
CN113029971B (zh) 一种作物冠层氮素监测方法及系统
CN102855485B (zh) 一种小麦抽穗的自动检测方法
KR101822410B1 (ko) 작물 영상을 사용한 작물 생육 진단 장치
CN102636486B (zh) 一种棉花三真叶期的自动检测方法
CN104949981A (zh) 一种棉花五真叶期自动检测方法及系统
CN106326822A (zh) 车道线检测的方法及装置
CN106483135A (zh) 基于机器视觉的复杂背景下玉米粒检测识别装置及方法
US10192112B2 (en) Plot gap identification
KR20200020209A (ko) 포장 단위별 다분광 영상 히스토그램 패턴 분석을 통한 작물 생육 모니터링 장치
Zhou et al. Autonomous detection of crop rows based on adaptive multi-ROI in maize fields.
CN102288776B (zh) 玉米植株生长速率测量方法
US8538695B2 (en) System and method for analyzing trees in LiDAR data using views
CN112541383B (zh) 杂草区域的识别方法和装置
CN102360428B (zh) 一种玉米三叶期与七叶期的自动检测方法
CN110781865A (zh) 一种农作物生长控制系统
Li et al. Image processing for crop/weed discrimination in fields with high weed pressure
CN110580495A (zh) 一种梨树叶面积与叶面炭疽病病斑数自动化分析方法
CN113963255A (zh) 一种基于田垄颜色特征的垄上作业机械导航线识别方法
CN113807129A (zh) 一种作物区域识别方法、装置、计算机设备和存储介质
CN114625114A (zh) 一种基于机器视觉的地面喷施系统行进路径规划方法
UA126999C2 (uk) Визначення смуги захоплення дії із зібраних машиною оброблених даних

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21821009

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 08.05.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 21821009

Country of ref document: EP

Kind code of ref document: A1