Navigation method and system for green cutting robot

Abstract

The invention discloses a green cutting robot navigation method and system, relates to the field of green cutting robot navigation, is used for improving the navigation precision of a green cutting robot, and comprises the following steps: acquiring different types of environmental information in front; the different types of environment information comprise image data and radar point cloud data; determining target crop path information independently identified according to the respective environment information; the reliability scores of the target crop path information independently identified according to the respective environment information are evaluated respectively; comparing the spatial geometric consistency of the target crop path information independently identified according to the respective environment information and the respective reliability score, and quantifying the overall credibility of the current environment information; when the total credibility is greater than or equal to a credibility threshold value, navigation is performed according to different types of environment information; when the total credibility is smaller than a credibility threshold value, navigation is carried out according to internal dead reckoning.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of mowing robot navigation, and in particular to a mowing robot navigation method and system. BACKGROUND

[0002] In the scenario of autonomous navigation of mowing robots in farmland, the navigation system usually adopts a combination of satellite positioning and local environment perception modules for navigation. However, some farmland surfaces often have subtle undulations that are difficult to detect with the naked eye, and the soil compaction degree of local areas is not uniform. When the mowing robot travels in such fields, the contact between its tracks or wheels and the ground will change. For example, when driving through a relatively soft and wet soil, the wheels may experience slight skidding; when encountering a locally compacted hard area, the robot's body may produce a slight pitch or roll angle change, resulting in a transient deviation between the robot's actual ground speed and the system's instructions. At the same time, the change in the robot's attitude also makes the height and inclination angle of the local perception sensors (such as stereo vision cameras and laser radars) carried by the robot relative to the ground no longer constant, which introduces potential error accumulation for subsequent fine navigation.

[0003] Furthermore, in dense areas, the leaves of adjacent plants may block each other, making it difficult to identify individual plant features, resulting in difficulty in accurately extracting clear "passage" boundaries in the vision processing process; in sparse areas, the gaps between plants are too large, and even rows are broken, making the concept of "row" ambiguous, and the system may misjudge or lose the path. Similarly, the echo signals of the laser radar are also affected, and the echo point cloud in dense areas may be too dense to distinguish individual plants; in sparse areas, there may be too few echo points to construct a complete row structure. This unevenness in crop growth makes the local perception data itself inherently uncertain, further exacerbating the difficulty of the navigation system in fine adjustment.

[0004] In addition, the shape, color, and height of weeds and other non-target plants in the field may be similar to those of the target crop (green feed corn). The mowing robot may misidentify these weeds or volunteer plants as part of the target crop, or consider them as interference. This leads to the robot making incorrect lateral adjustments and deviating from the true corn planting row. The laser radar may also receive echoes from these weeds, confusing them as corn plants, further interfering with the judgment of the real crop boundary.

[0005] Finally, the working characteristics of the robot itself and external environmental factors will also have a cumulative effect on the performance of the sensors during long periods of operation. For example, a large amount of particulate matter will gradually adhere and accumulate on the lens protection cover of the stereo vision camera and the emission and receiving window of the laser radar carried by the robot, causing the data quality obtained by the local perception system to continue to decline. This means that even if the robot is in a relatively ideal crop growth area, its navigation system may not be able to obtain sufficient reliable local environmental information due to the decline in the "vision" of the sensors, thereby affecting its judgment and fine adjustment ability of the row center, and ultimately may cause the robot to deviate from the preset path, and even cause damage to the crops. SUMMARY

[0006] The present application provides a green cutting robot navigation method for improving the navigation accuracy of the green cutting robot.

[0007] In a first aspect, to solve the above technical problems, the present application provides a green cutting robot navigation method, comprising: obtaining different types of front environmental information; the different types of environmental information include image data and radar point cloud data; determining target crop path information independently recognized according to respective environmental information, and respectively evaluating the reliability scores of the target crop path information independently recognized according to respective environmental information; comparing the spatial geometric consistency of the target crop path information independently recognized according to respective environmental information and the respective reliability scores, and quantifying the overall credibility of the current environmental information; when the overall credibility is greater than or equal to a credibility threshold, navigating according to different types of environmental information; when the overall credibility is less than the credibility threshold, navigating according to internal dead reckoning.

[0008] Optionally, respectively evaluating the reliability scores of the target crop path information independently recognized according to respective environmental information comprises: in the case of image data as the environmental information, performing binaryzation processing on the image data; based on the binaryzation processed image data, extracting edges in the image data; identifying and fitting a plurality of line segments from the edges; the line segments are used to indicate the possible boundaries of the crops; and calculating the image reliability score according to the number of line segments, parallelism, length of the plurality of line segments, and edge definition of the image.

[0009] Optionally, calculating the image reliability score according to the number of line segments, parallelism, length of the plurality of line segments, and edge definition of the image comprises: calculating the image reliability score according to an image reliability score calculation formula; the image reliability score calculation formula is: F1=w1*(S / S MAX )+w2*L+w3*Q;wherein F1 represents the image reliability score, w1, w2, w3 represent weight coefficients; S represents the number of effective line segments; S MAXrepresents the expected maximum number of line segments; L represents the average line segment parallelism; Q represents the average edge sharpness; the average line segment parallelism can be determined by calculating the standard deviation of the angles between all fitted line segments; the average edge sharpness is determined by calculating the average gradient value of the crop row edge region.

[0010] Optionally, the reliability scores of the target crop path information independently identified according to the respective environment information are respectively evaluated, including: in the case that the environment information is radar point cloud data, pre-processing the radar point cloud data; the pre-processing includes: performing noise point filtering processing, and removing points below a certain height threshold according to the installation height of the laser radar and the ground slope information to exclude ground points; identifying crop point clusters in the radar point cloud data; performing straight line fitting on the center points or boundary points of the point clusters to identify the spatial boundary lines formed by the crop stems; and calculating a radar reliability score according to the integrity of the clustering result, the linearity of the fitted straight line, and the uniformity of the crop row spacing.

[0011] Optionally, the radar reliability score is calculated according to the integrity of the clustering result, the linearity of the fitted straight line, and the uniformity of the crop row spacing, including: calculating the radar reliability score according to a radar reliability score calculation formula; the radar reliability score calculation formula includes: F2=w4*(J / J MAX )+w5*N+w6*H; wherein F2 represents the radar reliability score, w4, w5, and w6 represent weight coefficients; J represents the effective cluster number; J MAX represents the expected maximum cluster number; N represents the average fitting linearity; H represents the inverse of the standard deviation of the crop row spacing; the average fitting linearity is determined by the root mean square error of the fitting residual.

[0012] Optionally, the spatial geometric consistency of the target crop path information independently identified according to the respective environment information is compared, including: sampling within the target crop path information independently identified according to the respective environment information to obtain a series of corresponding point pairs; calculating the average Euclidean distance deviation between the point pairs; the average Euclidean distance deviation is represented as:

[0013]

[0014] wherein O L represents the average Euclidean distance deviation, N is the number of sampling points, X i is the horizontal coordinate of the sampling point, sqrt represents the square root function, fv(X i ) represents the functional representation of the target crop path information independently identified according to the image data; fl(X i ) represents the functional representation of the target crop path information independently identified according to the radar point cloud data.

[0015] Optionally, the overall credibility of the current environment information is quantified, comprising: determining the overall credibility according to the average Euclidean distance deviation; and the overall credibility satisfies the following relationship:

[0016]

[0017] F Z represents the overall credibility; O Lmax represents the maximum allowable deviation.

[0018] Optionally, the method further comprises: obtaining a virtual trajectory; the virtual trajectory is a smooth extension of the position and posture of the mowing robot when the overall credibility is less than the credibility threshold; following the virtual trajectory based on a steering instruction to navigate according to the dead reckoning; and the steering instruction is determined according to the deviation between the dead reckoning result and the virtual trajectory.

[0019] Optionally, the method further comprises: identifying a crop pixel region in the image data; identifying a crop point cluster in the radar point cloud data; projecting the crop point cluster to an image plane to obtain a projected point; for each projected point, determining whether it falls within a center confidence area of the corresponding crop pixel region; calculating a proportion of the number of projected points falling outside the center confidence area to the total number of projected points to obtain a projection deviation ratio; and when the projection deviation ratio exceeds a preset threshold, determining that there is a consistency deviation phenomenon, and correcting the overall credibility according to the consistency deviation phenomenon.

[0020] In a second aspect, the present application provides a mowing robot navigation system, which comprises:

[0021] An environment information acquisition module is configured to acquire different types of environment information in front; the different types of environment information comprise image data and radar point cloud data;

[0022] A path identification and evaluation module is configured to determine target crop path information independently identified according to respective environment information; and evaluate the reliability scores of the target crop path information independently identified according to respective environment information, respectively;

[0023] A perception information mutual verification module is configured to compare the spatial geometric consistency of the target crop path information independently identified according to respective environment information and the respective reliability scores, and quantify the overall credibility of the current environment information;

[0024] A navigation strategy adjustment module is configured to navigate according to different types of environment information when the overall credibility is greater than or equal to a credibility threshold.

[0025] When the overall credibility is less than the credibility threshold, navigate according to the dead reckoning.

[0026] Compared with the prior art, the present application has the following beneficial effects:

[0027] The mowing robot navigation method disclosed in the application obtains different types of front environment information (including image data and radar point cloud data), independently identifies target crop path information, and simultaneously evaluates respective reliability scores. On this basis, the spatial geometric consistency of the paths identified by different types of environment information and the respective reliability scores are compared, and the overall credibility of the current environment information is quantified. When the overall credibility reaches a preset threshold, the system navigates according to the multi-source environment information; when the overall credibility is lower than the threshold, the system switches to internal dead reckoning navigation.

[0028] This method effectively solves the problem that in the prior art, the complex and changeable farmland environment (such as ground undulation, uneven crop growth, non-target plant interference, and sensor performance degradation) leads to insufficient reliability of local perception data, thereby affecting navigation accuracy and robustness. Specifically, by introducing the fusion and mutual verification mechanism of multi-source environment information (image and radar point cloud), the application can overcome the limitations of a single sensor in harsh environments (such as light changes, dust, and crop shading), improve the comprehensiveness and accuracy of environmental perception. By independently evaluating the reliability of each sensor data and combining spatial geometric consistency for quantification, the system can judge the quality of the current perception information in real time, avoiding navigation errors caused by low-quality data. In addition, when the external perception information is unreliable, the system smoothly switches to internal dead reckoning navigation, ensuring the continuity and safety of the robot operation, and avoiding the interruption of the operation or damage to the crops caused by the failure of perception. In summary, the method of the application significantly improves the navigation robustness, reliability and adaptability of the mowing robot in complex farmland environments, ensuring the efficiency and accuracy of agricultural operations. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 is a mowing robot navigation method flowchart provided by an embodiment of the application;

[0030] Figure 2 is another mowing robot navigation method flowchart provided by an embodiment of the application;

[0031] Figure 3 is a mowing robot navigation system structure diagram provided by an embodiment of the application. DETAILED DESCRIPTION

[0032] The technical solutions in the present application will be described clearly and completely in the present application in combination with the drawings. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. The components of the present application described and shown in the drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work belong to the scope of protection of the present application.

[0033] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0034] The navigation method of the green cutting robot provided by the embodiments of the present application will be described and explained in detail below through the following specific embodiments.

[0035] With reference to Figure 1 The present application provides a navigation method of a green cutting robot, comprising the following steps:

[0036] S1, acquiring different types of environment information in front.

[0037] Among them, the different types of environment information include image data and radar point cloud data.

[0038] As a possible implementation manner, the green cutting robot can be equipped with a visual camera and a laser radar. The system can acquire image data in front through the visual camera of the green cutting robot, and acquire radar point cloud data in front through the laser radar.

[0039] Among them, the visual camera and the laser radar can be installed in front of the green cutting robot vehicle body and calibrated accurately to ensure that the data can be accurately converted into the coordinate system of the green cutting robot itself.

[0040] S2, determining the target crop path information independently recognized according to the respective environment information; and respectively evaluating the reliability score of the target crop path information independently recognized according to the respective environment information.

[0041] Among them, the target crop path information refers to the crop row center line or boundary line indicated by the robot that should be followed according to the environment information. For example, in a corn field, the target crop path information can be the geometric representation of the inter-row passage formed by the corn plants.

[0042] where the reliability score is a quantitative assessment of the quality of the target crop path information identified independently for each type of environmental information. The higher the score, the more reliable and accurate the path information is.

[0043] I. The process of independently identifying target crop path information based on image data can include:

[0044] Image preprocessing: First, perform color space conversion (e.g., from RGB to HSV) on the original image and apply filtering operations such as Gaussian blur to reduce image noise.

[0045] Crop region extraction: Utilize a specific range of the green channel in the HSV color space (e.g., H value between 30-90, S value between 40-255, V value between 30-255) to binarize the image, thereby preliminarily separating out potential crop regions.

[0046] Edge and line detection: Apply the Canny edge detection algorithm to the binarized image to extract all significant edges in the image. Subsequently, utilize the Hough Transform or RANSAC algorithm to identify and fit multiple straight lines or curve segments from these edges, which represent the possible boundaries of crop rows, obtaining the target crop path information independently identified based on image data.

[0047] II. The process of independently identifying target crop path information based on radar point cloud data can include:

[0048] Point cloud filtering and ground removal: First, filter out noise points from the original point cloud data (e.g., through statistical filtering or radius filtering). Since the mowing robot is mainly concerned with crop rows, the system will remove points below a certain height threshold (e.g., 0.1 meters) based on the installation height of the lidar and the ground slope information, to exclude ground points.

[0049] Crop point clustering and row identification: Apply a density-based clustering algorithm (e.g., DBSCAN) to the remaining point cloud data to cluster points belonging to the same plant or clump of crops. Subsequently, perform straight line fitting (e.g., least squares or RANSAC) on the center points or boundary points of these point clusters to identify the spatial boundary lines formed by crop stems, obtaining the target crop path information independently identified based on radar point cloud data.

[0050] Further, the system will calculate the image reliability score based on the number of identified line segments, their parallelism, length, and continuity in the image. The radar reliability score is determined based on the completeness of the clustering results, the linearity of the fitted straight lines, and the uniformity of the crop row spacing.

[0051] S3, compare the spatial geometric consistency and respective reliability scores of the target crop path information independently identified according to respective environmental information, and quantify the overall trustworthiness of the current environmental information.

[0052] wherein the spatial geometric consistency refers to the matching degree of the target crop path information independently identified by different types of sensors (such as image and radar) in spatial position and shape. The higher the consistency, the more consistent the perception results of different sensors on the same scene.

[0053] wherein the overall trustworthiness is a quantitative index of the overall trustworthiness of the current navigation data after considering the reliability scores of different environmental information and the spatial geometric consistency therebetween.

[0054] As a possible implementation manner, the system can sample within the target crop path information independently identified according to respective environmental information to obtain a series of corresponding point pairs; calculate the average Euclidean distance deviation between the point pairs; and determine the overall trustworthiness based on the average Euclidean distance deviation.

[0055] wherein the overall trustworthiness satisfies the following relationship:

[0056]

[0057] F Z represents the overall trustworthiness; O Lmax represents the maximum allowed deviation. F1 represents the image reliability score. F2 represents the radar reliability score. O L represents the average Euclidean distance deviation.

[0058] S4, when the overall trustworthiness is greater than or equal to the trustworthiness threshold, navigate according to different types of environmental information.

[0059] wherein the trustworthiness threshold can be set as needed and is not limited.

[0060] As a possible implementation manner, the system can use a high-gain proportional-integral-derivative (PID) controller or a model predictive control (MPC) algorithm, take the lateral deviation and heading deviation between the current position of the robot and the target path as input, and calculate accurate steering instructions and speed instructions to navigate the robot.

[0061] For example, the steering instruction can be expressed as:

[0062] Steering instruction = high lateral deviation K_p + high heading deviation K_d

[0063] wherein K_p and K_d are higher control gains.

[0064] S5, when the overall confidence is less than the confidence threshold, navigating according to dead reckoning.

[0065] The dead reckoning is a navigation method based on the kinematic model of the robot itself and the data of internal sensors (such as encoders, inertial measurement units IMU), which is used as a backup or supplementary navigation method when external perception data is unreliable.

[0066] As a possible implementation, the system can obtain a virtual trajectory and follow the virtual trajectory based on a steering instruction to navigate according to dead reckoning.

[0067] The steering instruction is determined according to the deviation between the dead reckoning result and the virtual trajectory.

[0068] The virtual trajectory is a smooth extension of the position and pose of the mowing robot when the overall confidence is less than the confidence threshold,

[0069] As another possible implementation, the system can obtain a virtual trajectory and follow the virtual trajectory based on a steering instruction to navigate according to dead reckoning when the duration that the overall confidence is less than the confidence threshold is less than a duration threshold.

[0070] The duration threshold can be set as needed. For example, it can be 3 seconds.

[0071] In some embodiments, when the duration that the overall confidence is less than the confidence threshold is greater than the duration threshold, it indicates that the robot has been unable to obtain any reliable local environment information, and the system will trigger an emergency shutdown procedure or send an alarm to a remote operator through a wireless communication module to request human intervention to prevent further path deviation or potential crop damage.

[0072] The mowing robot navigation method of the present application significantly improves the navigation robustness and accuracy of the robot in complex farmland environments by introducing reliability evaluation and spatial geometric consistency comparison of multi-source environmental information, and dynamically adjusting the navigation strategy according to the quantified overall credibility. Traditional navigation methods often rely too much on a single sensor or simple fusion, and when sensor data is disturbed or quality decreases, it is easy to cause navigation failure. For example, in the scenarios of sparse crops, weed interference or sensor contamination, a single vision or laser radar system may not be able to provide accurate path information. The present application can more comprehensively and accurately judge the overall credibility of the current environmental information by independently identifying and evaluating the reliability of image data and radar point cloud data, and further mutually verifying through spatial geometric consistency. When the credibility is high, the multi-source perception data is fully utilized for accurate navigation; when the credibility is low, the internal dead reckoning mode is switched to avoid navigation deviation caused by false perception data. This adaptive navigation strategy enables the mowing robot to maintain stable and reliable operation capability in various complex and variable farmland conditions, effectively solving the problem of insufficient navigation robustness in the prior art, and improving the operation efficiency and safety.

[0073] In one possible design, as shown in Figure 2 To evaluate the reliability score of the target crop path information independently identified according to each environmental information respectively, the present application can further include the following steps:

[0074] S101, in the case of environmental information being image data, the image data is subjected to binaryzation processing.

[0075] The purpose of binaryzation processing is to convert the pixel points in the image into only two possible values (usually black and white) according to their gray value or color value, thereby simplifying the image information and highlighting the contrast between the crops and the background.

[0076] As one possible implementation, the system can set a global or local threshold value, and set the pixels in the image higher than the threshold value as white (representing crops or foreground), and set the pixels lower than the threshold value as black (representing background).

[0077] S102, based on the image data subjected to binaryzation processing, edges in the image data are extracted.

[0078] Wherein, edge extraction aims to identify regions in the image where brightness or color changes significantly, which usually correspond to the boundaries of objects.

[0079] For example, classic edge detection algorithms such as Canny operator, Sobel operator or Prewitt operator can be employed to identify edge information in the image. These algorithms determine edge locations by computing the gradient of the image pixels, resulting in a clear crop boundary profile.

[0080] S103, identify and fit multiple line segments from the edges.

[0081] wherein the line segments are indicative of possible boundaries of the crop.

[0082] As a possible implementation, the system can utilize line detection algorithms such as Hough Transform to identify and fit continuous line segments from the discrete edge pixels. These fitted line segments are considered as potential boundary representations of crop rows or individual crop stalks.

[0083] In this way, the complex crop image information can be abstracted into a series of geometric line segments, facilitating subsequent analysis and processing.

[0084] S104, calculate the image reliability score according to the number of line segments, parallelism, length of the line segments, and edge sharpness of the image.

[0085] wherein the number of line segments reflects the richness of the identified crop boundaries; the parallelism measures the alignment degree between the line segments, usually crop rows are approximately parallel; the length indicates the integrity of the identified crop boundaries; and the edge sharpness of the image reflects the image quality and the explicitness of the crop boundaries. These parameters, taken together, can quantify the reliability of the image data in indicating crop paths.

[0086] As a possible implementation, the system can calculate the image reliability score according to the image reliability score calculation formula;

[0087] The image reliability score calculation formula is:

[0088] F1 = w1 * (S / S MAX ) + w2 * L + w3 * Q

[0089] wherein F1 represents the image reliability score, w1, w2, w3 represent weight coefficients; S represents the number of effective line segments; S MAX represents the expected maximum number of line segments; L represents the average line segment parallelism; Q represents the average edge sharpness; the average line segment parallelism can be determined by calculating the standard deviation of the included angle between all fitted line segments; and the average edge sharpness is determined by calculating the average gradient value of the crop row edge region.

[0090] By the technical scheme, an effective method for evaluating the reliability of target crop path information based on image data can be provided for the green cutting robot. The method can accurately identify the crop boundary from the complex farmland environment and convert it into quantifiable line segment features through image preprocessing, feature extraction and geometric fitting. By comprehensively considering the number, parallelism, length of the line segment and the edge definition of the image, the image reliability score can be objectively and accurately calculated, thereby providing a reliable basis for subsequent navigation decision. This helps to improve the navigation accuracy and robustness of the green cutting robot under different light, crop growth conditions and environmental complexity, and avoid navigation deviation or wrong decision caused by poor image information quality

[0091] In a possible design, in order to respectively evaluate the reliability scores of the target crop path information independently identified according to respective environment information, the application further includes the following steps:

[0092] S201, in the case of environment information being radar point cloud data, the radar point cloud data is preprocessed.

[0093] The preprocessing includes: noise point filtering processing, and removing points below a certain height threshold according to the installation height of the laser radar and the ground slope information to exclude ground points.

[0094] Specifically, when the system perceives an external environment fluctuation event, such as a sudden drop in air temperature or a sudden change in sunlight intensity, it will immediately intercept the power data sequence associated with the fluctuation event before and after the fluctuation event from the historical running data of the intelligent environment control device.

[0095] S202, identify the crop point cluster in the radar point cloud data.

[0096] The crop point cluster refers to a dense point set formed by a single or multiple crop stems and their leaves in the point cloud.

[0097] As a possible implementation, the identification of the crop point cluster can be realized by various clustering algorithms, such as DBSCAN (Density-Based Spatial Clustering Application with Noise) algorithm, K-means algorithm or region growing algorithm based on connectivity. These algorithms can aggregate points belonging to the same crop together to form independent crop point clusters according to the spatial density and distance relationship of the points.

[0098] S203, straight line fitting is performed on the center points or boundary points of the point clusters to identify the spatial boundary lines formed by the crop stems.

[0099] The spatial boundary lines usually represent the direction of the crop row.

[0100] As a possible implementation, the straight line fitting can employ least square method, RANSAC algorithm or other robust fitting methods to extract a representative straight line model from the discrete point clusters. The center points or boundary points of the point clusters are selected for fitting, depending on the specific crop morphology and fitting strategy, aiming to accurately represent the spatial position and direction of the crop rows.

[0101] S204, according to the integrity of the clustering result, the linearity of the fitted straight line and the uniformity of the crop row spacing, calculate the radar reliability score.

[0102] Among them, the integrity of the clustering result can measure the matching degree of the number of identified crop point clusters and the expected number of crop rows, for example, by calculating the ratio of the number of effective clusters to the expected maximum number of clusters. The linearity of the fitted straight line reflects the consistency of the fitted straight line and the actual crop rows, for example, which can be evaluated by calculating the root mean square error of the fitting residual. The uniformity of the crop row spacing is determined by analyzing the distance distribution between adjacent fitted straight lines, for example, calculating the inverse of the standard deviation of the row spacing, the smaller the standard deviation, the better the uniformity. These indicators comprehensively reflect the clarity and reliability of the crop row features in the radar point cloud data, thereby quantifying the radar reliability score.

[0103] As a possible implementation, the system can calculate the radar reliability score according to the radar reliability score calculation formula;

[0104] The radar reliability score calculation formula includes:

[0105] F2=w4*(J / J MAX )+w5*N+w6*H;

[0106] Wherein, F2 represents the radar reliability score, w4, w5, w6 represent the weight coefficient; J represents the effective cluster number; J MAX represents the expected maximum cluster number; N represents the average fitting linearity; H represents the inverse of the standard deviation of the crop row spacing; the average fitting linearity is determined by the root mean square error of the fitting residual.

[0107] The scheme of the present application designs special preprocessing, crop point cluster identification, spatial boundary line fitting and multi-dimensional reliability evaluation process by aiming at the characteristics of radar point cloud data, effectively solves the problem of insufficient accuracy of traditional general evaluation method in processing radar data. Specifically, noise point filtering and ground point removal ensure data purity, making subsequent crop identification more accurate. The identification of crop point cluster and the fitting of spatial boundary line can efficiently extract the key geometric information of crop row from radar data. Further, by comprehensively considering the integrity of the clustering result, the linearity of the fitted straight line and the uniformity of the crop row spacing, the quality and reliability of the crop path information identified from the radar data can be comprehensively and objectively evaluated, thereby providing a more reliable basis for subsequent navigation decision.

[0108] Through the above technical scheme, the accuracy and robustness of the reliability evaluation of the crop path information in the radar point cloud data by the mowing robot can be significantly improved. This targeted evaluation method enables the robot to more accurately judge the usability of the current radar perception data, avoiding navigation deviation caused by poor quality of radar data. Especially in environments where image data may be limited due to insufficient light and dust, accurate evaluation of radar reliability score can ensure that the robot can still navigate based on reliable perception information, thereby improving the adaptability and work efficiency of the mowing robot in complex farmland environments and reducing the risk of misoperation.

[0109] In some preferred embodiments, the following is illustrated by a specific example. Assume that a mowing robot is working in a farmland, and its laser radar is continuously obtaining the point cloud data in front. When the robot enters an area where the crops are sparsely grown or there are some weeds, the traditional general reliability evaluation method may not be able to accurately judge the reliability of the radar data. However, using the scheme of the present application, first, the radar point cloud data is preprocessed, for example, by statistical filtering to remove sporadic noise points, and according to the installation height of the laser radar, the ground points below 0.1 meters are removed to ensure that only crop and obstacle points are retained. Subsequently, the DBSCAN clustering algorithm is used to identify the crop point clusters in the point cloud. For example, if there are three rows of crops expected, the system will try to identify three main point clusters. Then, the center points of each identified crop point cluster are fitted with straight lines to obtain three spatial boundary lines representing the direction of the crop rows. Finally, the system calculates the radar reliability score. For example, if only two complete crop point clusters are identified (the completeness of the clustering result is low), the fitted straight lines have large bends (poor linearity), or the spacing between adjacent crop rows fluctuates greatly (poor uniformity), the calculated radar reliability score will be low. This low radar reliability score will be integrated with the reliability scores of other environmental information, and may cause the robot to switch to an internal dead reckoning navigation mode to avoid navigation errors caused by unreliable radar data, thereby ensuring the safety and efficient operation of the robot in complex environments.

[0110] In a possible design, in order to compare the spatial geometric consistency of the target crop path information independently identified according to respective environmental information, the present application further includes the following steps:

[0111] S301, sampling within the target crop path information independently identified according to respective environmental information to obtain a series of corresponding point pairs.

[0112] Among them, after obtaining the target crop path information independently identified according to the image data and the target crop path information independently identified according to the radar point cloud data, the two pieces of path information need to be compared in space. For this purpose, sampling can be performed along these path information. The sampling can adopt a uniform sampling method, that is, a series of sampling points are selected at fixed intervals in the length direction of the path; or adopt adaptive sampling, that is, increase the sampling density in the area where the curvature of the path changes greatly, so as to more accurately capture the geometric features of the path. Through sampling, a series of corresponding point pairs can be obtained, each of which contains a point from the image path information and a corresponding point from the radar path information.

[0113] S302, calculating the average Euclidean distance deviation between the point pairs.

[0114] The average Euclidean distance deviation is represented as:

[0115]

[0116] wherein O L represents the average Euclidean distance deviation, N is the number of sampling points, X i is the horizontal coordinate of the sampling point, sqrt represents the square root function, fv(X i ) represents the functional representation of the target crop path information identified independently according to the image data; fl(X i ) represents the functional representation of the target crop path information identified independently according to the radar point cloud data.

[0117] The scheme of the present application realizes the quantitative evaluation of the spatial geometric consistency of the two different source path information by sampling the target crop path information independently identified from the image data and the radar point cloud data and calculating the average Euclidean distance deviation. When the crop paths independently identified from the two types of sensor data highly coincide in space, the Euclidean distance deviation between the corresponding point pairs will be smaller, so that the value of the average Euclidean distance deviation O_L is also smaller. Conversely, if there is a significant spatial difference between the two paths, the average Euclidean distance deviation O_L will be larger. This quantitative method provides an objective index for judging whether the paths identified by different sensor data confirm each other, thereby providing a key input for subsequent overall credibility quantification.

[0118] Through the above technical scheme, an accurate and quantifiable method can be provided to evaluate the spatial geometric consistency between the target crop paths identified from different types of environmental information (such as image data and radar point cloud data). This quantitative evaluation avoids subjective judgment, enabling the mowing robot to make more accurate judgments on the reliability of the current environmental information based on objective data. Thus, in subsequent quantification of the overall credibility of the current environmental information, the advantages of different sensor data can be more effectively integrated, improving the robustness and accuracy of navigation decisions, especially in complex farmland environments where a single sensor data may have limitations or uncertainties.

[0119] In a possible design, in order to navigate according to the internal dead reckoning, the present application further comprises:

[0120] S401, acquiring a virtual trajectory.

[0121] wherein the virtual trajectory is a smooth extension according to the position and pose of the mowing robot when the overall credibility is less than the credibility threshold.

[0122] Specifically, the virtual trajectory can be understood as an expected or smoothed path calculated by the mowing robot system based on its historical motion data and current state of internal sensors (such as odometer, inertial measurement unit IMU, etc.) when external environment perception information is unreliable. The virtual trajectory aims to provide a temporary and relatively stable navigation reference to cope with the situation of insufficient or inaccurate external perception data. Among them, the smooth extension of the virtual trajectory means that when the external perception data is unreliable, the robot does not completely stop or move randomly, but based on its recent valid motion trend and attitude information, a smooth and continuous path is predicted and generated to maintain the continuity of navigation.

[0123] S402, follow the virtual trajectory based on the steering instruction to navigate according to the internal dead reckoning.

[0124] Among them, the steering instruction is determined according to the deviation of the dead reckoning result and the virtual trajectory. This means that the system will continuously monitor the difference between the current position and attitude of the mowing robot obtained by internal dead reckoning and the preset virtual trajectory. Once the deviation is detected, the corresponding steering instruction is generated to guide the robot to adjust its motion direction and attitude to make it as close as possible to the virtual trajectory. The dead reckoning result is the current position and attitude of the robot calculated in real time based on the kinematic model of the robot itself and internal sensor data (such as wheel encoder, gyroscope, etc.).

[0125] The scheme of the present application introduces a virtual trajectory as a navigation reference when external environment information is unreliable, and generates a steering instruction based on the deviation of the internal dead reckoning result and the virtual trajectory, so that the mowing robot can still maintain basic path tracking ability in the absence of high-credibility external perception data. This mechanism ensures that the robot can maintain a certain autonomous navigation ability even in complex or adverse environmental conditions, avoiding complete stagnation or deviation from the predetermined path. Through the smooth extension of the virtual trajectory, the robot can smoothly transition from multi-sensor fusion navigation to internal dead reckoning navigation, ensuring the continuity and stability of navigation.

[0126] Through the above technical scheme, when the overall credibility of external environment information is low, the mowing robot can smoothly transition to the navigation mode based on internal dead reckoning. This method avoids navigation interruption or path loss caused by unreliable external perception data, significantly improves the robustness and operation continuity of the robot in complex environments. Through the smooth extension of the virtual trajectory and the deviation-based steering instruction, the robot can achieve more stable and predictable autonomous navigation, effectively reducing the risk of operation and ensuring the smooth progress of mowing operation.

[0127] However, in practical applications, even if the macro spatial geometry is highly consistent, subtle calibration errors between sensors, dynamic pose drifts, or local environmental complexities (e.g., crop occlusions, lighting variations) can still cause inconsistencies in image data and radar point cloud data at a detailed level. Such inconsistencies, which are not fully identified and quantified, can lead to inaccurate overall trustworthiness assessments of current environmental information, thereby affecting the robustness and accuracy of navigation decisions.

[0128] To this end, the present application further proposes a navigation method for a green cutting robot, the method further comprising:

[0129] S501, identifying a crop pixel region in the image data.

[0130] Wherein, the identification of the crop pixel region in the image data can employ image segmentation techniques, such as a deep learning-based semantic segmentation model trained to accurately distinguish between crop and non-crop regions in the image, thereby identifying the crop pixel region in the image data.

[0131] S502, identifying a crop point cluster in the radar point cloud data.

[0132] Wherein, the identification of the crop point cluster in the radar point cloud data can utilize clustering algorithms, such as DBSCAN or K-means, to process the radar point cloud data to identify point clusters representing individual or multiple crop stalks.

[0133] S503, projecting the crop point cluster to the image plane to obtain projected points.

[0134] As a possible implementation, the system can convert the point coordinates in the three-dimensional radar point cloud to a two-dimensional image coordinate system to project the crop point cluster to the image plane to obtain the projected points.

[0135] S504, for each projected point, determining whether it falls within a center confidence region of the corresponding crop pixel region.

[0136] Wherein, the center confidence region can be understood as a core region within the crop pixel region, and its size and shape can be adjusted according to actual application requirements, for example, it can be defined as a region extending a certain number of pixels outward from the geometric center of the crop pixel region, or obtained from the original crop pixel region through morphological operations (such as erosion). This is aimed at excluding areas of crop edge blur or uncertainty and focusing on the consistency of the main part of the crop.

[0137] S505, calculating the proportion of the number of projected points falling outside the center confidence region to the total number of projected points to obtain a projection deviation ratio.

[0138] S506, when the projection deviation ratio exceeds the preset threshold value, it is judged that there is a consistency deviation phenomenon, and the overall credibility is corrected according to the consistency deviation phenomenon.

[0139] The preset threshold value can be determined according to experience or by experiment, and is used to define the acceptable inter-sensor consistency range.

[0140] Once the consistency deviation phenomenon is detected, the previously calculated overall credibility can be corrected accordingly. The correction method can include but is not limited to: reducing the calculated overall credibility score, or increasing the weight of dead reckoning navigation in subsequent navigation decisions, to reflect the decrease in reliability of the current multi-sensor fusion information.

[0141] Through the above technical solutions, the application can effectively improve the robustness and reliability of the mowing robot navigation system in complex environments. Specifically, by introducing cross-modal consistency verification of image and radar data, the system can more finely evaluate the quality of multi-sensor fusion information, making up for the shortcomings of relying only on spatial geometric consistency and independent reliability score evaluation. When there is a slight deviation or local inconsistency in sensor data, this solution can quickly detect and quantify such deviations, and then correct the overall credibility, avoiding navigation risks caused by inaccurate perception information. Thus, even in the case of limited sensor performance or complex and variable environments, the system can make more prudent and accurate navigation strategy adjustments, ensuring that the mowing robot can safely and efficiently work along the target crop path, significantly improving the accuracy and safety of navigation.

[0142] In some preferred embodiments, the following is described by a specific example. Suppose the mowing robot is navigating in a farmland, and its image sensor and laser radar sensor may have a slight relative pose drift due to vibration or temperature change after long-term operation, resulting in a slight deviation in the spatial perception of the same crop row by the two sensors.

[0143] When the robot obtains the front environment information:

[0144] First, the image data is processed to identify the crop pixel area, for example, by using a semantic segmentation model to accurately outline the boundary of the crop row.

[0145] At the same time, the radar point cloud data is processed to identify the crop point cluster, for example, by using a clustering algorithm to identify a set of point clouds representing crop stems.

[0146] Then, these crop point clusters are projected onto the image plane to generate a series of projected points.

[0147] The system checks whether each projected point falls into the center confidence region of the corresponding crop pixel region in the image. For example, if a crop row is identified in the image, the center confidence region is defined as the center 50% width range of the crop row pixel region.

[0148] If it is found that a considerable portion of the projected points (e.g., more than a preset threshold of 10%) fall outside the center confidence region, it indicates that there is inconsistency between the perception of the crop location by the image sensor and the radar sensor, i.e., there is a consistency deviation phenomenon.

[0149] Based on this consistency deviation phenomenon, the system will correct the overall confidence previously calculated. For example, even if the original overall confidence score is high, it will be reduced by a preset penalty value, or directly multiplied by a correction coefficient less than 1.

[0150] Thus, the corrected overall confidence more accurately reflects the true reliability of the current multi-sensor fusion information. If the corrected overall confidence is lower than the confidence threshold, the system will switch to dead reckoning for navigation in time, thereby avoiding navigation errors caused by relying on inaccurate fusion information, ensuring that the robot can continue to work safely and stably until the sensor consistency is restored or the environmental conditions improve.

[0151] As shown in FIG. 1, the present application also provides a navigation system for a green chopping robot. The system comprises: Figure 3 An environment information acquisition module for acquiring different types of environment information in front; the different types of environment information include image data and radar point cloud data;

[0152] A path identification and evaluation module for determining target crop path information independently identified according to respective environment information; and respectively evaluating the reliability scores of the target crop path information independently identified according to respective environment information;

[0153] A perception information mutual verification module for comparing the spatial geometric consistency of the target crop path information independently identified according to respective environment information and the respective reliability scores, and quantifying the overall confidence of the current environment information;

[0154] A navigation strategy adjustment module for navigating according to different types of environment information when the overall confidence is greater than or equal to a confidence threshold;

[0155] When the overall confidence is less than the confidence threshold, navigating according to dead reckoning.

[0156]

[0157] ​The embodiments of the present application further provide a computer readable storage medium. All or part of the processes in the above method embodiments can be instructed by a computer program to relevant hardware to complete, the program can be stored in the above computer readable storage medium, and the program can include the processes of the above method embodiments when executed. The computer readable storage medium can be an internal storage unit of the task execution apparatus (including the data sending end and / or the data receiving end) of any of the above embodiments, such as a hard disk or a memory of the task execution apparatus. The above computer readable storage medium can also be an external storage device of the terminal device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device. Further, the above computer readable storage medium can include both the internal storage unit of the task execution apparatus and the external storage device. The above computer readable storage medium is used to store the above computer program and other programs and data required by the task execution apparatus. The above computer readable storage medium can also be used to temporarily store data that has been output or will be output.

[0158] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0159] When the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present application essentially or the parts that make contributions to the prior art or all or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium, includes several instructions to make a device (which can be a single-chip microcomputer, a chip, etc.) or a processor execute all or part of the steps of the methods of the embodiments of the present application. The above-mentioned storage medium includes: a U disk, a mobile hard disk, a ROM, a RAM, a magnetic disk or an optical disk, and various media that can store program codes.

[0160] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any change or replacement within the technical scope disclosed in the present application should be covered in the protection scope of the present application.

Claims

1. A navigation method for a green-cutting robot, characterized in that, include: Acquire different types of environmental information ahead; these different types of environmental information include image data and radar point cloud data. Determine the target crop path information that is independently identified based on its respective environmental information, and evaluate the reliability score of the target crop path information that is independently identified based on its respective environmental information. By comparing the spatial geometric consistency and reliability scores of target crop path information independently identified based on their respective environmental information, the overall credibility of the current environmental information is quantified. When the overall credibility is greater than or equal to the credibility threshold, navigation is performed based on the different types of environmental information; When the overall confidence level is less than the confidence level threshold, navigation is performed based on internal dead reckoning.

2. The navigation method for a green-cutting robot according to claim 1, characterized in that, The evaluation of the reliability scores of target crop path information independently identified based on their respective environmental information includes: When the environmental information is image data, the image data is binarized. Based on the binarized image data, edges are extracted from the image data. Multiple line segments are identified and fitted from the edges; these line segments are used to indicate possible boundaries of the crop. The image reliability score is calculated based on the number of line segments, parallelism, length, and edge sharpness of the image.

3. The navigation method for a green-cutting robot according to claim 2, characterized in that, The step of calculating the image reliability score based on the number of line segments, parallelism, length, and edge sharpness of the image includes: Calculate the image reliability score according to the image reliability score calculation formula; The formula for calculating the image reliability score is: F1=w1*(S / S MAX )+w2*L+w3*Q Where F1 represents the image reliability score, w1, w2, and w3 represent weighting coefficients; S represents the number of valid line segments; S MAX L represents the expected maximum number of line segments; Q represents the average line segment parallelism; the average edge sharpness is determined by calculating the standard deviation of the angles between all fitted line segments; the average edge sharpness is determined by calculating the average gradient value of the crop row edge region.

4. The navigation method for a green-cutting robot according to claim 1, characterized in that, The reliability scores of target crop pathway information independently identified based on their respective environmental information were evaluated separately, including: When the environmental information is radar point cloud data, the radar point cloud data is preprocessed; the preprocessing includes: noise point filtering, and removing points below a specific height threshold based on the installation height of the lidar and the ground slope information to exclude ground points; Identify crop clusters in the radar point cloud data; By fitting straight lines to the center or boundary points of these point clusters, the spatial boundary lines formed by crop stems can be identified. The radar reliability score is calculated based on the completeness of the clustering results, the linearity of the fitted line, and the uniformity of the crop row spacing.

5. The navigation method for a green-cutting robot according to claim 4, characterized in that, The radar reliability score is calculated based on the completeness of the clustering results, the linearity of the fitted line, and the uniformity of the crop row spacing, including: Calculate the radar reliability score according to the radar reliability score calculation formula; The formula for calculating the radar reliability score includes: F2w4*(J / J MAX +w5*N+w6*H Where F2 represents the radar reliability score, w4, w5, and w6 represent weighting coefficients; J represents the number of effective clusters; J MAX The expected maximum number of clusters is represented by N; the average linearity of the fit is represented by H; the average linearity of the fit is represented by the reciprocal of the standard deviation of the crop row spacing; the average linearity of the fit is determined by the root mean square error of the fit residuals.

6. The navigation method for a green-cutting robot according to claim 1, characterized in that, The comparison of the spatial geometric consistency of target crop path information independently identified based on their respective environmental information includes: Sampling is performed within the target crop path information independently identified based on their respective environmental information to obtain a series of corresponding point pairs; Calculate the average Euclidean distance deviation between the point pairs; The average Euclidean distance deviation is expressed as: Among them, O L This represents the average Euclidean distance deviation, where N is the number of sampling points, and X... i Here, x is the x-coordinate of the sampling point, sqrt represents the square root function, and fv(X) i fl(X) represents the function representation of target crop path information independently identified from image data; i ) represents a function that independently identifies the target crop path information based on radar point cloud data.

7. The navigation method for a green-cutting robot according to claim 6, characterized in that, The overall credibility of the quantified current environmental information includes: The overall confidence level is determined based on the average Euclidean distance deviation. The overall credibility satisfies the following relationship: F Z Indicates overall credibility; O Lmax This indicates the maximum permissible deviation.

8. The navigation method for a green-cutting robot according to claim 1, characterized in that, The navigation based on internal dead reckoning includes: Obtain a virtual trajectory; the virtual trajectory is a smooth extension based on the position and posture of the cutting robot when the overall confidence level is less than the confidence level threshold; The virtual trajectory is followed based on the turning command to navigate according to internal dead reckoning; the turning command is determined based on the deviation between the dead reckoning result and the virtual trajectory.

9. A navigation method for a green-cutting robot according to claim 1, characterized in that, The method further includes: Identify crop pixel regions in the image data; Identify crop clusters in the radar point cloud data; The crop point clusters are projected onto the image plane to obtain the projection points; For each projection point, determine whether it falls within the central confidence region of the corresponding crop pixel area; The projection deviation ratio is obtained by calculating the proportion of projection points that fall outside the central confidence region to the total number of projection points. When the projection deviation ratio exceeds a preset threshold, it is determined that there is a consistency deviation phenomenon, and the overall credibility is corrected based on the consistency deviation phenomenon.

10. A navigation system for a green-cutting robot, characterized in that, The system includes: An environmental information acquisition module is used to acquire different types of environmental information ahead; the different types of environmental information include image data and radar point cloud data. The path identification and evaluation module is used to determine the target crop path information independently identified based on their respective environmental information; and to evaluate the reliability score of the target crop path information independently identified based on their respective environmental information. The perception information mutual verification module is used to compare the spatial geometric consistency of target crop path information independently identified based on their respective environmental information and their respective reliability scores to quantify the overall credibility of the current environmental information. The navigation strategy adjustment module is used to perform navigation based on the different types of environmental information when the overall credibility is greater than or equal to the credibility threshold. When the overall confidence level is less than the confidence level threshold, navigation is performed based on internal dead reckoning.

Images