Method for quantifying weed space and evaluating work load in orchard road area

Abstract

The application provides a method for quantifying weed space and evaluating work load in orchard road area, which comprises the following steps: collecting real-time scene image in front of the orchard road, and performing image enhancement processing to obtain green vegetation candidate area; inputting the scene image in front of the orchard road into a road segmentation model to output a road region mask; performing road constraint screening on the green vegetation candidate area based on the road region mask, and extracting the weed candidate area in the road; performing morphological processing and connected domain screening on the weed candidate area in the road, and obtaining the weed mask in the road to calculate the global weed coverage rate; constructing a geometric self-adaptive grid, calculating the grid cell coverage rate, spatial heterogeneity and aggregation index to obtain the work load index. The application can reduce the interference of the crown layer vegetation outside the road, shadow, soil texture and complex light on weed identification, improve the stability of weed extraction and spatial quantification in the orchard road area, and provide a basis for variable weeding and intelligent operation decision.

Description

Technical Field

[0001] This application belongs to the fields of agricultural robots, machine vision, intelligent weeding and precision agriculture, and specifically relates to a method for spatial quantification of weeds and assessment of workload in orchard road areas. It is applicable to application scenarios such as weed identification, spatial distribution quantification, variable weeding decision-making and workload assessment in orchard weeding robots, orchard plant protection robots and intelligent orchard operation equipment in road areas. Background Technology

[0002] Orchard roads and inter-row areas are crucial passageways for mechanized operations, plant protection, weeding, and transportation in orchards. As orchard management evolves towards intelligent, refined, and low-input methods, accurate identification and quantitative assessment of weeds within orchard road areas are becoming essential prerequisites for intelligent weeding equipment and variable-rate operation systems. Continuous weed growth within roads not only reduces the effective passage space for robots and agricultural machinery but also increases operational resistance, affects driving stability, and may lead to unreasonable allocation of plant protection or weeding intensity.

[0003] Existing methods for identifying weeds in orchards or farmland mainly include thresholding methods based on color indices, image processing methods based on color space transformation, and object detection or semantic segmentation methods based on deep learning. While these methods can achieve green vegetation target recognition to a certain extent, they still have limitations in orchard road scenarios. On the one hand, orchard road areas are typically affected by factors such as tree canopy occlusion, shadow variations, strong light reflection, bare soil background, and vegetation on the outer canopy of the road. Relying solely on color thresholds or vegetation indices can easily lead to false positives and false negatives. On the other hand, orchard roads have obvious geometric boundaries and operational constraints. The green canopy outside the road and the weeds inside the road may have similar color features in the image. Without introducing road area constraints, it is easy to misjudge the green background outside the road as weeds inside the road.

[0004] Furthermore, existing weed identification methods mostly focus on determining the presence of weeds or calculating overall coverage, with less emphasis on further analyzing the local spatial distribution, patch aggregation, and spatial heterogeneity of weeds within roads. However, in actual variable-based weeding operations, a single global coverage rate is insufficient to reflect the differences in operational intensity across different areas within roads. Weeds in orchard roads often exhibit patchy, aggregated, and uneven distribution characteristics, and the coverage, number of patches, and aggregation degree in different local areas directly affect weeding intensity, pesticide demand, machinery workload, and operational priority.

[0005] Therefore, there is an urgent need to provide a method for spatial quantification of road-constrained weeds and assessment of workload in orchard road areas, so as to achieve stable extraction of weeds within roads, quantification of spatial structure, and comprehensive evaluation of workload under complex lighting, irregular road boundaries, and background interference conditions, and to provide a basis for variable weeding and intelligent operation decision-making in orchards. Summary of the Invention

[0006] The purpose of this application is to provide a method for spatial quantification of weeds and assessment of workload in orchard road areas. This method aims to address the problems in existing technologies where weed identification on orchard roads is easily affected by external green backgrounds, shadows, soil textures, and complex lighting, and it is difficult to further characterize the spatial distribution of weeds and the complexity of workload within the road. This method can stably extract weed targets within the road area under road area constraints and construct a geometrically adaptive mesh based on the actual road boundaries, thereby achieving a comprehensive assessment of weed coverage, spatial heterogeneity, aggregation degree, and workload.

[0007] Firstly, this application provides a method for quantifying weed space and assessing workload in orchard road areas, including: S1. Real-time acquisition of images of the scene ahead of the orchard road; S2. Perform image enhancement processing on the scene image in front of the orchard road based on green vegetation information to obtain green vegetation candidate areas; S3. Input the scene image in front of the orchard road into the road segmentation model built based on YOLOv11, and output the road area mask; S4. Based on the road area mask, green vegetation candidate areas are screened under road constraints, and weed candidate areas within the road are extracted based on the HSV color threshold and the ExG vegetation index threshold. S5. Perform morphological processing and connected component filtering on the candidate areas of weeds in the road to obtain a weed mask in the road, and calculate the global weed coverage rate based on the number of weed pixels in the road and the number of effective road area pixels. S6. Construct a geometrically adaptive mesh based on the actual boundary of the road area mask, and calculate the mesh cell coverage, spatial heterogeneity, and clustering index based on the geometrically adaptive mesh; S7. Based on the global weed coverage, spatial heterogeneity, and aggregation index, obtain the workload index and output the spatial quantification results of weeds in the orchard road area and the workload assessment results.

[0008] Preferably, the scene image in front of the orchard road includes the main body of the road, weeds inside the road, road boundaries, tree canopy background on both sides, bare soil area, shadow area and high reflectivity area.

[0009] Preferably, the image of the scene in front of the orchard road is subjected to image enhancement processing based on green vegetation information to obtain candidate areas of green vegetation, including: S21. Convert the scene image in front of the orchard road to the LAB color space and perform contrast-limited adaptive histogram equalization on the brightness channel. S22. Convert the equalized image to the HSV color space and adjust the gain of the saturation and value channels. S23. Calculate the ExG vegetation index based on the RGB channels. The ExG vegetation index is ExG=2G-RB, where R, G, and B represent the gray values ​​of the red, green, and blue channels of the scene image in front of the orchard road, respectively. S24. Based on the ExG vegetation index, the threshold segmentation method is used to obtain the initial green vegetation candidate region, and the initial green vegetation candidate region is normalized by morphological opening and closing operations to obtain the green vegetation candidate region.

[0010] Preferably, the threshold segmentation method uses the Otsu algorithm or a preset threshold.

[0011] Preferably, the road segmentation model based on YOLOv11 includes introducing a Squeeze-and-Excitation channel attention mechanism into the key feature layer of the YOLOv11 model.

[0012] Preferably, S4 includes: S41. Resample the road area mask to the same spatial resolution as the scene image in front of the orchard road; S42. Extract the candidate regions for green vegetation based on the HSV color threshold to obtain the first candidate mask for weeds; S43. Based on the ExG vegetation index threshold, the candidate areas of green vegetation are extracted to obtain the second weed candidate mask; S44. Perform an intersection operation on the road area mask, the first weed candidate mask, and the second weed candidate mask to obtain the weed candidate area within the road.

[0013] Preferably, S5 includes: S51. Use elliptical structural elements to perform morphological opening operations on candidate areas of weeds within the road to remove isolated noise points; S52. Perform morphological closing operations on candidate weed regions within roads that have undergone morphological opening operations to fill small pores inside weed patches and enhance patch connectivity. S53. Based on the connected component area threshold, invalid small patches with areas smaller than a preset threshold are removed to obtain the final road weed mask. , ,in, γ( ) represents the morphological opening operation. φ ( ) represents the morphological closing operation. Ψ ( ) indicates a noise removal operation based on a preset threshold; S54. Calculate the global weed coverage rate. ,in, N weed This indicates the number of pixels representing weeds along the road. N road Indicates the number of pixels in the effective road area. This represents the overall weed coverage rate.

[0014] Preferably, constructing a geometrically adaptive mesh based on the actual boundary of the road area mask includes: firstly, extracting the left and right boundaries of the road area mask; then dividing the road area into multiple longitudinal strip units along the longitudinal direction of the road; for each longitudinal strip unit, determining the effective road width based on the left and right boundaries of the road within that longitudinal strip unit; and then dividing the effective road width into multiple transverse mesh units according to a preset transverse ratio, thereby obtaining a geometrically adaptive mesh consistent with the actual geometric boundary of the road.

[0015] Preferably, the workload index Where a, b, and c represent the weighting coefficients of global weed coverage, coefficient of variation, and aggregation index, respectively. , For overall weed coverage, The coefficient of variation is 1. AI This is the aggregation index.

[0016] On the other hand, this application also provides a system for quantifying weed space and assessing workload in orchard road areas, including: The image acquisition module is used to acquire real-time images of the scene in front of the orchard road; The green vegetation enhancement module is used to perform image enhancement processing on the scene image in front of the orchard road based on green vegetation information to obtain green vegetation candidate areas. The road region segmentation module is used to input the scene image in front of the orchard road into the road segmentation model built based on YOLOv11 and output the road region mask; The road constraint weed extraction module is used to perform road constraint screening on green vegetation candidate areas based on the road area mask, and to extract road weed candidate areas based on HSV color threshold and ExG vegetation index threshold. The weed basic quantization module is used to perform morphological processing and connected component filtering on the candidate weed areas in the road to obtain a weed mask in the road, and calculate the global weed coverage rate based on the number of weed pixels in the road and the number of effective road area pixels. The geometrically adaptive mesh construction module is used to construct a geometrically adaptive mesh based on the actual boundary of the road area mask, and to calculate the mesh cell coverage, spatial heterogeneity, and clustering index based on the geometrically adaptive mesh. The workload assessment module is used to obtain the workload index based on the global weed coverage, spatial heterogeneity, and aggregation index, and output the spatial quantification results of weeds in the orchard road area and the workload assessment results.

[0017] Compared with the prior art, this application has the following beneficial effects: First, this application improves the separability between green weed targets and bare soil, shadows, high-reflectivity areas and other non-green backgrounds in orchard road scenes by enhancing green vegetation, thereby enhancing the stability of subsequent weed candidate region extraction.

[0018] Second, this application obtains a road area mask by constructing a road segmentation model based on YOLOv11, and uses the road area mask as a spatial constraint for weed extraction, which can reduce false detections caused by vegetation on the outer canopy of the road and green targets in the background.

[0019] Third, this application uses HSV color threshold and ExG (excessive greenness index) vegetation index threshold to extract candidate areas of weeds in the road, which can take into account both large areas of green weeds and small-scale fragmented weed patches, thus improving the completeness of weed identification in the road.

[0020] Fourth, this application constructs a geometrically adaptive mesh based on the actual road boundary, avoiding the problem of mismatch between the traditional rectangular mesh and the perspective shape of the orchard road, and can more accurately reflect the local spatial distribution characteristics of weeds within the road.

[0021] Fifth, this application quantifies roadside weeds at multiple scales by using global weed coverage, grid cell coverage, spatial heterogeneity, aggregation index, and workload index, which can provide data support for variable weeding, workload allocation, and intelligent operation decision-making.

[0022] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures and processes shown in the description and the accompanying drawings. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the method for quantifying weed space and assessing workload in orchard road areas provided in this application embodiment.

[0025] Figure 2 This is a schematic diagram of the orchard road image acquisition and data processing process provided in the embodiments of this application.

[0026] Figure 3 This is a schematic diagram of the image enhancement processing flow based on green vegetation information provided in the embodiments of this application.

[0027] Figure 4 This is a schematic diagram of the orchard road area segmentation model structure based on the improved YOLOv11 provided in the embodiments of this application.

[0028] Figure 5 This is a schematic diagram of the road constraint weed extraction and basic quantification process provided in the embodiments of this application.

[0029] Figure 6 This is a schematic diagram of weed spatial distribution quantification and workload assessment based on road geometry adaptive grid provided in the embodiments of this application. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0031] Please refer to Figure 1 This application provides a method for quantifying weed space and assessing workload in orchard road areas, including: S1. Real-time acquisition of images of the scene ahead of the orchard road; S2. Perform image enhancement processing on the scene image in front of the orchard road based on green vegetation information to obtain green vegetation candidate areas; S3. Input the scene image in front of the orchard road into the road segmentation model built based on YOLOv11, and output the road area mask; S4. Based on the road area mask, green vegetation candidate areas are screened under road constraints, and weed candidate areas within the road are extracted based on the HSV color threshold and the ExG vegetation index threshold. S5. Perform morphological processing and connected component filtering on the candidate areas of weeds in the road to obtain a weed mask in the road, and calculate the global weed coverage rate based on the number of weed pixels in the road and the number of effective road area pixels. S6. Construct a geometrically adaptive mesh based on the actual boundary of the road area mask, and calculate the mesh cell coverage, spatial heterogeneity, and clustering index based on the geometrically adaptive mesh; S7. Based on the global weed coverage, spatial heterogeneity, and aggregation index, obtain the workload index and output the spatial quantification results of weeds in the orchard road area and the workload assessment results.

[0032] In a preferred embodiment, in step S1, the image of the scene in front of the orchard road includes information such as the main body of the road, weeds within the road, road boundaries, tree canopy background on both sides, bare soil areas, shaded areas, and highly reflective areas. For example... Figure 2 This includes an orchard operation platform, a positioning unit, a vehicle-mounted image acquisition device, an image acquisition path, and examples of orchard road images. The examples of orchard road images represent different forward-looking road scenes obtained under the same acquisition process, illustrating that the acquired images contain information such as the main road structure, road boundaries, weeds within the road, and the background of tree canopies on both sides. The vehicle-mounted image acquisition device, installed on an orchard weeding robot, plant protection robot, or other intelligent operation platform, acquires images of the scene ahead of the orchard road in real time. Preferably, the image acquisition device is a vehicle-mounted high-definition camera, or one or more of an RGB camera, industrial camera, and depth camera. The images acquired by the image acquisition device can be single frames or image frames from a continuous video stream.

[0033] It should be noted that in step S2, weeds within the orchard road scene typically appear as green vegetation targets. However, the canopy vegetation outside the road, shadows, bare soil texture, and changes in lighting can affect the stable identification of green targets. Therefore, this application performs image enhancement processing on the scene image in front of the orchard road based on green vegetation information to improve the separability between green weed areas and non-green backgrounds.

[0034] In a preferred embodiment, such as Figure 3 The diagram shows a flowchart of image enhancement processing based on green vegetation information, including a green enhanced grayscale response map, a green enhanced pseudo-color map, and the final green vegetation enhancement result obtained by enhancing the original image. Image enhancement processing based on green vegetation information is performed on the scene image in front of the orchard road to obtain green vegetation candidate regions, including: S21. Convert the image of the scene in front of the orchard road to the LAB color space, and perform contrast-limited adaptive histogram equalization on the luminance channel to reduce the effects of local shadows and uneven lighting. Let the original luminance channel be... L The enhanced luminance channel is L′ Then we have: L′=f CLAHE (L) ,in, f CLAHE ( ) This indicates a contrast-limited adaptive histogram equalization operator.

[0035] S22. Convert the equalized image to the HSV color space and adjust the gain of the saturation and value channels to enhance the color salience of the green vegetation area.

[0036] S23. Calculate the ExG vegetation index based on the RGB channels to enhance the response difference between green vegetation targets and non-green backgrounds; ExG vegetation index ExG=2G-RB ,in, R, G, B These represent the grayscale values ​​of the red, green, and blue channels of an image depicting a scene in front of the orchard road. These values ​​can be obtained directly from the original acquired image or by converting the image back to the RGB color space after LAB brightness equalization and HSV gain adjustment. In a preferred embodiment, R, G, and B represent the 8-bit grayscale values ​​of the red, green, and blue channels in the same image. The ExG vegetation index is used to enhance the response difference of green vegetation relative to a non-green background, forming a green response map or a green enhancement response map. In an optional embodiment, a green enhancement expression can also be constructed: G boost =αG-βR-βB ,in, G boost This indicates the enhanced green response. α This represents the gain coefficient of the green channel. β This represents the suppression coefficients for the red and blue channels.

[0037] S24. Based on the ExG vegetation index, a threshold segmentation method is used to obtain initial green vegetation candidate regions. Morphological opening and closing operations are then performed to normalize these initial green vegetation candidate regions, resulting in the final green vegetation candidate regions. Specifically, the threshold segmentation method can be the Otsu algorithm or a preset threshold. The threshold segmentation method based on the ExG vegetation index is used to obtain the initial green vegetation candidate regions. M 0 Morphological opening and closing operations were used to analyze the initial candidate areas for green vegetation. M 0After standardization, candidate areas for green vegetation were obtained. M v ,Right now M v = φ ( γ ( M 0 In this process, γ() represents the morphological opening operation, and φ() represents the morphological closing operation. The above processing can improve the target response intensity of green vegetation, target separability, and spatial coherence of candidate regions.

[0038] In a preferred embodiment, in step S3, as Figure 4 As shown, a road segmentation model based on YOLOv11 is used to segment the scene image in front of the orchard road, and the road region mask is output. M road The road area mask is used to define the effective operating area for subsequent weed extraction and spatial quantization.

[0039] Preferably, a Squeeze-and-Excitation channel attention mechanism is introduced into the key feature layers of the YOLOv11 model to enhance the representation of road structure features and suppress background interference such as canopy vegetation, shadows, soil texture, and weeds. Let the feature map input to the YOLOv11 model be... F First, channel description vectors are obtained through global average pooling. z Then, channel weights s are obtained through learnable parameter mapping and nonlinear activation functions, and finally, the recalibrated feature map is obtained. F ',Right now s = σ ( W 2 δ ( W 1 GAP ( F ))) F ′= s ⊙ F ,in, GAP ( ) indicates global average pooling. W 1 and W 2 Indicates learnable parameters, δ ( )express ReLU Activation function σ( ) ⊙ represents the Sigmoid activation function, and ⊙ represents channel-wise multiplication.

[0040] Through the above processing, the road segmentation model can stably extract effective road areas even when orchard roads are obscured by weeds, have discontinuous boundaries, and are subject to background interference, providing a spatial boundary basis for subsequent road constraint weed extraction.

[0041] In a preferred embodiment, in step S4, as Figure 5 As shown, the process of filtering candidate green vegetation areas based on the road area mask and extracting candidate weed areas within the road based on the HSV color threshold and ExG vegetation index threshold includes: S41, Mask the road area M road Resample to the same spatial resolution as the scene image in front of the orchard road to ensure that the road area mask and the scene image in front of the orchard road have a pixel-level correspondence; S42. Extract the candidate regions for green vegetation based on the HSV color threshold to obtain the first candidate mask for weeds. M HSV It should be noted that the first weed candidate mask is obtained by screening green vegetation candidate areas using HSV color thresholds. Pixels that meet the H, S, and V threshold ranges are marked as candidate weed pixels; otherwise, they are marked as background pixels. The HSV color thresholds are determined based on the color distribution differences between green weed targets and bare soil, shadows, and the outer canopy background in the orchard road scene, through representative sample statistics, manual observation, and experimental optimization. In a preferred embodiment, the HSV thresholds are set as follows: H ∈ [28, 90], S ∈ [40, 255], V ∈ [40, 255], used to extract green vegetation candidate areas.

[0042] S43. Based on the ExG vegetation index threshold, extract the green vegetation candidate region to obtain the second weed candidate mask. M ExG It should be noted that the second weed candidate mask is obtained by thresholding the ExG vegetation index map. Pixels with ExG values ​​greater than or equal to a set threshold are marked as candidate weed pixels. The ExG vegetation index threshold is used to determine whether a pixel has a significant green vegetation response. In a preferred embodiment, the ExG vegetation index threshold is set to 20. This value was determined through experimental debugging and visual inspection of representative orchard road images, with the principle of setting it to preserve fragmented weed areas within the road as much as possible while suppressing soil, shadows, and false green noise.

[0043] S44. Perform an intersection operation on the road area mask, the first weed candidate mask, and the second weed candidate mask to obtain the weed candidate area within the road. M weed ,Right now Mweed =M road ∩M HSV ∩M ExG .

[0044] In step S4, by introducing a road area mask constraint, this application can effectively eliminate the influence of tree canopies, roadside weeds, and other green background targets outside the road on the identification of weeds inside the road. By combining the HSV color threshold and the ExG vegetation index threshold, both large areas of green weeds and small-scale fragmented weed patches can be taken into account, thereby improving the stability and completeness of weed extraction inside the road.

[0045] In a preferred embodiment, in step S5, it should be noted that the number of pixels representing weeds within the road is obtained from a road weeds mask. Specifically, after obtaining the road weeds mask, the number of pixels with a value of 1 or the number of images of the scene ahead of the road in the mask are counted as the number of road weeds pixels. The number of effective road area pixels is obtained from a road area mask, that is, the number of pixels belonging to the road area in the road area mask is counted and recorded as the effective road area pixel count. Since the road area mask has been resampled to the same spatial resolution as the original image, the two can be directly calculated in the same pixel coordinate system.

[0046] Specifically, candidate areas for weeds within roads. M weed The process involves morphological processing and connected component filtering, and calculating the global weed coverage rate based on the number of weed pixels within the road and the number of effective road area pixels. Specifically, this includes: S51, performing morphological opening operations on candidate weed regions within the road using elliptical structuring elements to remove isolated noise points; S52, performing morphological closing operations on the candidate weed regions within the road after morphological opening operations to fill small holes within weed patches and enhance patch connectivity; and S53, removing invalid small patches with areas smaller than a preset threshold based on a connected component area threshold to obtain the final road weed mask. ,Right now ,in, γ (·) indicates the morphological opening operation. φ (·) represents the morphological closing operation. Ψ (·) indicates noise removal operation based on a preset threshold; S54, global weed coverage. ,in, N weed This indicates the number of pixels representing weeds along the road. N road Indicates the number of pixels in the effective road area. The global weed coverage rate is used to characterize the overall weed coverage level in the orchard road area.

[0047] In a preferred embodiment, in step S6, it should be noted that, as Figure 6 As shown, orchard roads in front-view images typically exhibit a perspective shape that converges at the far end and expands at the near end, and their boundaries are not equivalent to the rectangular boundaries of the entire image. If the traditional rectangular grid division method is used directly, it is easy to cause the grid cells to mismatch with the actual road boundaries, thereby affecting the local weed coverage and spatial distribution statistics.

[0048] Therefore, this application is based on road area masking. M road The actual boundaries are used to construct a geometrically adaptive mesh. Specifically, the left and right boundaries of the road area mask are first extracted; then the road area is divided into multiple longitudinal strip units along the longitudinal direction of the road; for each longitudinal strip unit, the effective road width is determined according to the left and right boundaries of the road within the longitudinal strip unit; then the effective road width is divided into multiple transverse mesh units according to a preset transverse ratio, thus obtaining a geometrically adaptive mesh that is consistent with the actual geometric boundaries of the road.

[0049] For the effective grid cell in the i-th row and j-th column, its grid cell coverage is... Represented as: ,in, This represents the number of weed pixels in the i-th row and j-th column of the grid cell. This represents the number of valid road pixels in the grid cell at row i and column j. It's important to note that a valid grid cell refers to a grid cell that, after geometrically adaptive meshing, intersects with the road region mask and contains valid road pixels. In other words, only grid cells containing road region pixels participate in the calculation of grid cell coverage, spatial heterogeneity, and subsequent workload; grid cells completely outside the road region or with zero valid road pixels are not included in the statistics.

[0050] Furthermore, the mean coverage of each effective grid cell is calculated. and standard deviation And using the coefficient of variation Characterizing spatial heterogeneity, ,in, The larger the value, the more uneven the spatial distribution of weeds within the road, and the more obvious the differences in local operations.

[0051] Meanwhile, the aggregation index is calculated based on the area of ​​weed patches within the road. AI : ,in, A max This indicates the area of ​​the largest weed patch. This represents the sum of the areas of all weed patches. AI Used to characterize whether weeds within roads are concentrated in large patches, when AIWhen the size is larger, it indicates that the weeds are more concentrated. AI When the value is smaller, it indicates that the weeds are more dispersed or fragmented.

[0052] In a preferred embodiment, in step S7, the spatial quantification results of weeds in the orchard road area include weed cover within the road, global weed coverage, coverage of each grid cell, spatial heterogeneity, aggregation index, and the corresponding spatial distribution map or quantification matrix. The workload assessment results include the workload index (WLI), workload level, and optional variable weeding operation recommendations, such as weeding intensity level, pesticide application level, mechanical weeding frequency, or operation priority.

[0053] Specifically, based on the overall weed coverage rate Coefficient of variation and clustering index AI Calculate the workload index ,Right now Where a, b, and c represent the weighting coefficients of global weed coverage, coefficient of variation, and aggregation index, respectively. Preferably, the settings can be configured according to the orchard weeding needs. It can also be adjusted according to different operating equipment, weeding methods, and management objectives.

[0054] Workload Index This is used to comprehensively characterize the complexity of weeding operations in orchard road areas. Global weed coverage reflects the overall weed load, and spatial heterogeneity (coefficient of variation) is also considered. The coverage rate reflects local differences in workload, while the aggregation index reflects the concentration of weed patches. By weighting and combining these three factors, a more comprehensive assessment of workload can be obtained than that based on a single coverage rate.

[0055] In an alternative implementation, the workload index can be used. The road area is divided into low-load, medium-load, and high-load zones, and corresponding variable-based weeding operation suggestions are output. For example, when... When the load is less than or equal to the first threshold, it is determined to be a low-load area; when When the load is between the first and second thresholds, it is determined to be a medium load region; when Areas with a weeding load greater than or equal to the second threshold are identified as high-load areas. The operational recommendations may include weeding intensity level, pesticide application level, mechanical weeding frequency, or operational priority.

[0056] This application also provides a system for quantifying weed space and assessing workload in orchard road areas, used to implement the aforementioned method for quantifying weed space and assessing workload in orchard road areas. The system includes: The image acquisition module is used to acquire real-time images of the scene in front of the orchard road; The green vegetation enhancement module is used to perform image enhancement processing on the scene image in front of the orchard road based on green vegetation information to obtain green vegetation candidate areas. The road region segmentation module is used to input the scene image in front of the orchard road into the road segmentation model built based on YOLOv11 and output the road region mask; The road constraint weed extraction module is used to perform road constraint screening on green vegetation candidate areas based on the road area mask, and to extract road weed candidate areas based on HSV color threshold and ExG vegetation index threshold. The weed basic quantization module is used to perform morphological processing and connected component filtering on the candidate weed areas in the road to obtain a weed mask in the road, and calculate the global weed coverage rate based on the number of weed pixels in the road and the number of effective road area pixels. The geometrically adaptive mesh construction module is used to construct a geometrically adaptive mesh based on the actual boundary of the road area mask, and to calculate the mesh cell coverage, spatial heterogeneity, and clustering index based on the geometrically adaptive mesh. The workload assessment module is used to obtain the workload index based on the global weed coverage, spatial heterogeneity, and aggregation index, and output the spatial quantification results of weeds in the orchard road area and the workload assessment results.

[0057] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for quantifying weed space and assessing workload in orchard road areas, characterized in that, include: S1. Real-time acquisition of images of the scene ahead of the orchard road; S2. Perform image enhancement processing on the scene image in front of the orchard road based on green vegetation information to obtain green vegetation candidate areas; S3. Input the scene image in front of the orchard road into the road segmentation model built based on YOLOv11, and output the road area mask; S4. Based on the road area mask, green vegetation candidate areas are screened under road constraints, and weed candidate areas within the road are extracted based on the HSV color threshold and the ExG vegetation index threshold. S5. Perform morphological processing and connected component filtering on the candidate areas of weeds in the road to obtain a weed mask in the road, and calculate the global weed coverage rate based on the number of weed pixels in the road and the number of effective road area pixels. S6. Construct a geometrically adaptive mesh based on the actual boundary of the road area mask, and calculate the mesh cell coverage, spatial heterogeneity, and clustering index based on the geometrically adaptive mesh; S7. Based on the global weed coverage, spatial heterogeneity, and aggregation index, obtain the workload index and output the spatial quantification results of weeds in the orchard road area and the workload assessment results.

2. The method for quantifying weed space and assessing workload according to claim 1, characterized in that, The image of the scene in front of the orchard road includes the main body of the road, weeds inside the road, road boundaries, tree canopy background on both sides, bare soil area, shadow area and high reflectivity area.

3. The method for quantifying weed space and assessing workload according to claim 1, characterized in that, Image enhancement processing based on green vegetation information is performed on the scene image in front of the orchard road to obtain candidate regions of green vegetation, including: S21. Convert the scene image in front of the orchard road to the LAB color space and perform contrast-limited adaptive histogram equalization on the brightness channel. S22. Convert the equalized image to the HSV color space and adjust the gain of the saturation and value channels. S23. Calculate the ExG vegetation index based on the RGB channels. The ExG vegetation index is ExG=2G-RB, where R, G, and B represent the gray values ​​of the red, green, and blue channels of the scene image in front of the orchard road, respectively. S24. Based on the ExG vegetation index, the threshold segmentation method is used to obtain the initial green vegetation candidate region, and the initial green vegetation candidate region is normalized by morphological opening and closing operations to obtain the green vegetation candidate region.

4. The method for quantifying weed space and assessing workload according to claim 3, characterized in that, The threshold segmentation method selected is the Otsu algorithm or a preset threshold.

5. The method for quantifying weed space and assessing workload according to claim 1, characterized in that, The road segmentation model based on YOLOv11 includes the introduction of a Squeeze-and-Excitation channel attention mechanism into the key feature layer of the YOLOv11 model.

6. The method for quantifying weed space and assessing workload according to claim 1, characterized in that, S4 include: S41. Resample the road area mask to the same spatial resolution as the scene image in front of the orchard road; S42. Extract the candidate regions for green vegetation based on the HSV color threshold to obtain the first candidate mask for weeds; S43. Based on the ExG vegetation index threshold, the candidate areas of green vegetation are extracted to obtain the second weed candidate mask; S44. Perform an intersection operation on the road area mask, the first weed candidate mask, and the second weed candidate mask to obtain the weed candidate area within the road.

7. The method for quantifying weed space and assessing workload according to claim 1, characterized in that, S5 include: S51. Use elliptical structural elements to perform morphological opening operations on candidate areas of weeds within the road to remove isolated noise points; S52. Perform morphological closing operations on candidate weed regions within roads that have undergone morphological opening operations to fill small pores inside weed patches and enhance patch connectivity. S53. Based on the connected component area threshold, invalid small patches with areas smaller than a preset threshold are removed to obtain the final road weed mask. , ,in, γ ( ) represents the morphological opening operation. φ ( ) represents the morphological closing operation. Ψ ( ) indicates a noise removal operation based on a preset threshold; S54. Calculate the global weed coverage rate. ,in, N weed This indicates the number of pixels representing weeds along the road. N road Indicates the number of pixels in the effective road area. This represents the overall weed coverage rate.

8. The method for quantifying weed space and assessing workload according to claim 1, characterized in that, Constructing a geometrically adaptive mesh based on the actual boundaries of the road area mask includes: firstly, extracting the left and right boundaries of the road area mask; then dividing the road area into multiple longitudinal strip units along the longitudinal direction of the road; for each longitudinal strip unit, determining the effective road width based on the left and right boundaries of the road within that longitudinal strip unit; and then dividing the effective road width into multiple transverse mesh units according to a preset transverse ratio, thereby obtaining a geometrically adaptive mesh consistent with the actual geometric boundaries of the road.

9. The method for quantifying weed space and assessing workload according to claim 1, characterized in that, Workload Index Where a, b, and c represent the weighting coefficients of global weed coverage, coefficient of variation, and aggregation index, respectively. , For overall weed coverage, The coefficient of variation is 1. AI This is the aggregation index.

10. A system for quantifying weed space and assessing workload in orchard road areas, characterized in that, include: The image acquisition module is used to acquire real-time images of the scene in front of the orchard road; The green vegetation enhancement module is used to perform image enhancement processing on the scene image in front of the orchard road based on green vegetation information to obtain green vegetation candidate areas. The road region segmentation module is used to input the scene image in front of the orchard road into the road segmentation model built based on YOLOv11 and output the road region mask; The road constraint weed extraction module is used to perform road constraint screening on green vegetation candidate areas based on the road area mask, and to extract road weed candidate areas based on HSV color threshold and ExG vegetation index threshold. The weed basic quantization module is used to perform morphological processing and connected component filtering on the candidate weed areas in the road to obtain a weed mask in the road, and calculate the global weed coverage rate based on the number of weed pixels in the road and the number of effective road area pixels. The geometrically adaptive mesh construction module is used to construct a geometrically adaptive mesh based on the actual boundary of the road area mask, and to calculate the mesh cell coverage, spatial heterogeneity, and clustering index based on the geometrically adaptive mesh. The workload assessment module is used to obtain the workload index based on the global weed coverage, spatial heterogeneity, and aggregation index, and output the spatial quantification results of weeds in the orchard road area and the workload assessment results.

Images