Obstacle detection method and unmanned vehicle

Abstract

The present invention discloses an obstacle detection method and an unmanned vehicle. The method relates to vehicle technology and autonomous driving, and includes: in response to a work vehicle detecting first obstacle data, matching the first obstacle data with second obstacle data; in response to a successful match between the first obstacle data and the second obstacle data, determining a fusion parameter based on the obstacle type in the first obstacle data; based on the fusion parameter, fusing the first obstacle shape in the first obstacle data with the second obstacle shape in the second obstacle data to obtain an initial obstacle shape; smoothing the initial obstacle shape to obtain a target obstacle shape; and updating the second obstacle data based on the target obstacle shape. The present invention solves the technical problem of low stability and accuracy in obstacle detection by work vehicles in related technologies.

Description

Technical Field

[0001] The present invention relates to the fields of vehicle technology and autonomous driving, and in particular to an obstacle detection method and an unmanned vehicle. Background Art

[0002] In application scenarios such as autonomous mining, operating vehicles often face complex and ever-changing environmental challenges. When dealing with small obstacles such as rocks and high ruts, obstacle detection methods in related technologies have significant limitations. Due to the small size of small obstacles in mining scenarios, factors such as vibrations generated by operating vehicles during driving and sensor noise can lead to unstable obstacle shape detection and frequent shape changes. This instability manifests itself as multiple close-knit small obstacles being mistakenly identified as a single obstacle, while a single obstacle is sometimes split into multiple obstacles, seriously affecting the autonomous driving performance of the operating vehicle. Unstable obstacle detection can cause the operating vehicle's autonomous driving system to make incorrect obstacle avoidance decisions, such as sudden braking, or inefficient path planning when navigating around small obstacles. This increases the time cost and energy consumption of mining operations and may even pose safety risks. In other words, the stability and accuracy of obstacle detection by operating vehicles in related technologies are low.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] The embodiments of the present invention provide an obstacle detection method and an unmanned vehicle, so as to at least solve the technical problem in the related art that a working vehicle has low stability and accuracy in detecting obstacles.

[0005] According to one aspect of an embodiment of the present invention, an obstacle detection method is provided, comprising: in response to a work vehicle detecting first obstacle data, matching the first obstacle data with second obstacle data, wherein the first obstacle data is obstacle data identified based on current perception data of the work vehicle, and the second obstacle data is obstacle data identified based on historical perception data of the work vehicle at historical times; in response to a successful match between the first obstacle data and the second obstacle data, determining a fusion parameter based on an obstacle type in the first obstacle data; fusing a first obstacle shape in the first obstacle data with a second obstacle shape in the second obstacle data based on the fusion parameter to obtain an initial obstacle shape; smoothing the initial obstacle shape to obtain a target obstacle shape; and updating the second obstacle data based on the target obstacle shape.

[0006] According to another aspect of an embodiment of the present invention, an unmanned vehicle is provided, comprising: a matching module for, in response to a work vehicle detecting first obstacle data, matching the first obstacle data with second obstacle data, wherein the first obstacle data is obstacle data identified based on perception data of the work vehicle at a current moment, and the second obstacle data is obstacle data of obstacles identified based on historical perception data of the work vehicle at historical moments; a determination module for, in response to a successful match between the first obstacle data and the second obstacle data, determining a fusion parameter based on an obstacle type in the first obstacle data; a fusion module for, based on the fusion parameter, fusing and smoothing a first obstacle shape in the first obstacle data with a second obstacle shape in the second obstacle data to obtain an initial obstacle shape; a smoothing module for smoothing the initial obstacle shape to obtain a target obstacle shape; and an updating module for updating the second obstacle data based on the target obstacle shape.

[0007] According to another aspect of an embodiment of the present invention, an electronic device is provided, including: a communication unit for communicating with a target vehicle; a memory storing an executable program; and a processor for running the program, wherein the method of each embodiment of the present invention is executed when the program is running.

[0008] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0009] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0010] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0011] According to another aspect of the embodiments of the present invention, a computer program is provided. When the computer program is executed by a processor, the methods in various embodiments of the present invention are implemented.

[0012] In an embodiment of the present invention, a more accurate and stable description of obstacle shapes is achieved by fusion of historical data and current data through smoothing. Specifically, obstacle data detected in real time by a work vehicle is matched with data from historical moments. Once a match is successful, fusion parameters are dynamically determined based on the obstacle type. This fusion parameter determination takes into account the physical properties of different obstacle types and their impact on the work vehicle's travel. Dynamic adjustment of the fusion parameters can adapt to the characteristics of different obstacle types. Based on the determined fusion parameters, the currently detected obstacle shape is fused with the historical shapes to generate a preliminary fused initial obstacle shape. This initial obstacle shape is then smoothed to eliminate potential noise and improve the quality of the shape description. This effectively offsets errors that may occur in single-moment detection and enhances the continuity and stability of the shape description. Based on the processed target obstacle shape, the historical obstacle data is updated, achieving iterative data improvement. This further improves the accuracy of subsequent obstacle detection by the work vehicle, thereby resolving the technical issues of low stability and accuracy in obstacle detection by work vehicles in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0014] Figure 1 is a flow chart of an obstacle detection method according to an embodiment of the present invention;

[0015] Figure 2 is a schematic diagram of an optional obstacle detection process according to an embodiment of the present invention;

[0016] Figure 3 is a schematic diagram of an optional obstacle smoothing process according to an embodiment of the present invention;

[0017] Figure 4 is a schematic diagram of an initial obstacle shape obtained by an optional fusion process according to an embodiment of the present invention;

[0018] Figure 5 is a schematic diagram of an optional process of obtaining multiple first projection points according to an embodiment of the present invention;

[0019] Figure 6 is a schematic diagram of an optional process of obtaining filtered vertices according to an embodiment of the present invention;

[0020] Figure 7 is a schematic diagram of an optional boundary line and projection line according to an embodiment of the present invention;

[0021] Figure 8 Schematic diagram of functional modules of an unmanned vehicle according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0023] It should be noted that the terms "first," "second," and the like in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the numbers used in this manner are interchangeable where appropriate so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatus.

[0024] According to one aspect of an embodiment of the present invention, a method for obstacle detection is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0025] Figure 1 is a flow chart of an obstacle detection method according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0026] Step S102 : In response to the work vehicle detecting the first obstacle data, matching the first obstacle data with the second obstacle data.

[0027] The first obstacle data is obstacle data identified based on the perception data of the working vehicle at the current moment, and the second obstacle data is obstacle data identified based on the historical perception data of the working vehicle at the historical moment.

[0028] The above-mentioned working vehicle may refer to an autonomous driving vehicle that performs working tasks in a specific environment. For example, it may be an autonomous driving mining truck in a mining scene. The working vehicle can also be determined according to actual needs and is not limited here.

[0029] The above-mentioned first obstacle data may refer to obstacle information identified based on the perception data of the working vehicle at the current moment. The first obstacle data may include data such as the type, size, position and shape of the obstacle, which can be determined according to actual needs and is not limited here.

[0030] The aforementioned second obstacle data may refer to obstacle information identified and recorded by the work vehicle at a historical moment, i.e., one or more moments prior. For example, if the work vehicle detects a rock ahead during its current driving process, the first obstacle data may be obtained. Furthermore, the work vehicle's system records may also contain information about rocks detected during previous driving, which can serve as the second obstacle data.

[0031] In an optional embodiment, in scenarios such as mining operations, work vehicles can be equipped with multiple sensors, such as lidar and cameras, to capture real-time information about the surrounding environment. The work vehicle's environmental perception equipment can process the output of these sensors, identify and extract obstacle information at the current moment, and form first obstacle data. Simultaneously, the work vehicle can retain obstacle information perceived at previous moments as second obstacle data for subsequent fusion reference. Next, based on the continuous perception data over a time series, obstacles identified at the current moment can be matched with those identified at previous moments. The matching process can include comparing multi-dimensional information such as location coordinates, size, and shape features to ensure matching accuracy. For example, candidate obstacles can first be selected by comparing the obstacle's center location coordinates and size range. The shape features of the candidate obstacles, such as angle changes and edge contours, are then further analyzed to determine the final matching result. During the matching process, feature fusion technology, such as obstacle feature extraction based on deep learning and improved matching algorithms, can also be used to improve the accuracy of obstacle data matching. It can quickly extract the key features of obstacles from massive perception data and find more similar matching objects in historical data. Even if the shape or position of the obstacle changes slightly, accurate matching can be achieved.

[0032] In the above process, data matching effectively combines the accuracy of real-time data with the stability of historical data. This is especially true for small obstacles in mining environments, such as rocks and wheel tracks. The long-term information contained in historical data can compensate for detection fluctuations in real-time data due to vehicle vibration or environmental interference, ensuring the stability of obstacle shape output. The matching process can identify which obstacle information is continuous and valid and which is newly appeared or no longer exists. This allows for selective data fusion, reducing computing resource consumption and improving obstacle detection efficiency. It also provides more reliable basic data for subsequent shape fusion and smoothing.

[0033] Step S104 : In response to the first obstacle data successfully matching the second obstacle data, determining a fusion parameter based on the obstacle type in the first obstacle data.

[0034] The above-mentioned obstacle types may refer to the classification of obstacles, such as stones, high ruts, curbs, etc. Different obstacle types may be handled with different strategies or parameters, which may be determined according to actual needs and are not limited here.

[0035] The aforementioned fusion parameters may refer to rules used to determine how to combine the first obstacle data and the second obstacle data to form a more accurate and stable description of the obstacle shape. Fusion parameters may include projection weights, smoothing coefficients, etc., and can be determined based on actual needs and are not limited here.

[0036] In an optional embodiment, after identifying the first and second obstacle data, the obstacle type in the first obstacle data can be accurately determined. Obstacle types may include, but are not limited to, static objects such as rocks, dynamic objects such as small animals, or specific environmental features such as wheel tracks. The identification process can utilize machine learning models, such as deep learning models, to intelligently classify obstacle types by analyzing multi-dimensional features such as obstacle images, size, location, and movement trends. When an obstacle is identified as a small obstacle in a mine, such as a rock or wheel track, fusion parameters can be automatically adjusted. These fusion parameters can take into account factors such as the obstacle's size, shape change rate, and environmental stability. For small obstacles, especially those that frequently appear in the vehicle's path, a higher weight can be assigned to the current perception data to reflect the newer obstacle's shape. Furthermore, to maintain consistency and stability in shape descriptions, the second projection parameters of the historical data can be set low to ensure that even if the obstacle's shape changes, the fused shape does not drastically change. The parameter adjustment strategy can undergo extensive verification and refinement to adapt to different types of obstacles and environmental conditions.

[0037] By dynamically adjusting fusion parameters based on obstacle type, the above process can flexibly determine the fusion ratio of new and old data according to the characteristics of different obstacles, thereby enhancing detection robustness while ensuring shape accuracy. For example, for small, static obstacles with minimal shape changes, current data is given a higher weight to promptly reflect the obstacle's new state; while for objects with easily changeable shapes, historical data is maintained at a certain weight to avoid sudden changes in shape caused by accidental errors.

[0038] Step S106 : Based on the fusion parameters, the first obstacle shape in the first obstacle data and the second obstacle shape in the second obstacle data are fused to obtain an initial obstacle shape.

[0039] The first obstacle shape mentioned above may refer to a shape description of an obstacle identified at the current moment, such as a set of vertex coordinates of a polygon, which may be determined according to actual needs and is not limited here.

[0040] The second obstacle shape mentioned above may refer to an obstacle shape description recorded based on historical moments.

[0041] The aforementioned initial obstacle shape may refer to an obstacle shape description generated by the obstacle data at the current moment and the historical moment after fusion processing, and may be in the form of a polygon and may include a series of vertex coordinates.

[0042] In an optional embodiment, before fusion, the two obstacle data sets can be preprocessed to ensure alignment of the first and second obstacle data sets in the spatial coordinate system. This can include operations such as coordinate transformation, rotation, and translation, so that the first and second obstacle data sets can be compared and fused in the same reference system. The first obstacle shape in the first obstacle data set can be fused with the second obstacle shape in the second obstacle data set. This fusion process can include merging and smoothing, and the fusion method can be based on weighted averaging, median, or other statistical methods to determine how the fused shape is maintained. For example, for each vertex position, the weighted average coordinate of each vertex in the two data sets can be calculated to generate the fused vertex position. Overlapping areas can also be processed to ensure shape consistency and integrity.

[0043] In this process, fusion combines obstacle information from both the current moment and historical moments to ensure the consistency and stability of obstacle shape descriptions. Even with changing environmental conditions, controlled by fusion parameters, the fused shape can still accurately reflect the obstacle's true state, reducing shape fluctuations caused by momentary environmental interference. Fusion of obstacle shapes at two different moments achieves a smooth transition in shape descriptions, avoiding sudden braking or additional path adjustments that could result from sudden shape changes, thereby improving the smoothness and safety of autonomous driving.

[0044] Step S108: Smoothing the initial obstacle shape to obtain the target obstacle shape.

[0045] The target obstacle shape mentioned above may refer to an improved obstacle shape description after smoothing and point number adjustment. The target obstacle shape can ensure the accuracy of the obstacle shape and the simplicity of the description, which facilitates real-time processing and decision-making of the vehicle.

[0046] In an optional embodiment, after the fusion process is completed, an initial obstacle shape is obtained that incorporates features from both the new and old data. This initial obstacle shape may appear rough due to noise or inconsistencies in the data fusion. Optionally, the fused initial obstacle shape can be preprocessed, such as removing duplicate points, outliers, or overly dense vertices to simplify the shape's complexity. Smoothing can be performed. This can include selecting an appropriate mathematical algorithm, such as a moving average, Gaussian filtering, Bezier curve fitting, or spline function interpolation, to smooth the obstacle's boundary contours, reducing sharp changes in the shape and making the shape appear more natural and more consistent with the object's actual appearance. For example, the moving average method smoothes the shape by calculating the local average position of obstacle edge points; the Gaussian filtering method uses Gaussian distribution weights to smooth point positions, thereby reducing the impact of random noise. The effectiveness of the smoothing process depends on the selection of specific parameters, such as window size, filter scale, and number of fits. These parameters can be modified based on the obstacle's size, shape complexity, and the real-time vehicle motion. An iterative approach can be employed to gradually adjust the parameters until a satisfactory smoothing effect is achieved while maintaining the integrity of the shape details. After smoothing, the smoothed obstacle shape can be evaluated to check for over-smoothing or shape distortion. If necessary, corrections can be made, such as restoring certain key feature points or locally adjusting the shape curve to ensure that the shape reflects the true state of the obstacle while meeting computational efficiency and smoothness requirements.

[0047] In the above process, smoothing effectively reduces the risk of incorrect decisions caused by shape data fluctuations or noise. The smoothed shape is more stable, ensuring the consistency of obstacle detection results, facilitating accurate obstacle avoidance strategies and improving autonomous driving safety. For the visual perception system, smoothed obstacle shapes are easier to identify and track, presenting clearer, more coherent outlines. This helps reduce the complexity of visual processing and improve recognition speed and accuracy, especially in complex terrain and lighting conditions such as mines.

[0048] Step S110: updating the second obstacle data based on the target obstacle shape.

[0049] In an optional embodiment, after obtaining the target obstacle shape after matching, fusion, and smoothing, the target obstacle shape is derived from a combination of current and historical data, resulting in a more accurate and stable result. The target obstacle shape information can be updated in the secondary obstacle data to reflect the updated obstacle status. Update strategies can include overwriting historical records, incremental updates, or fusing multiple detection results, depending on the data management mechanism and the emphasis placed on historical records. The update process can include the following steps: deleting the record corresponding to the target obstacle shape in the historical data; inserting or updating the complete target obstacle information in the database, including location coordinates, shape description, and type identification; and updating related metadata, such as the obstacle detection timestamp and confidence score, for subsequent use. The consistency of the obstacle data in the database can also be ensured by checking whether the updated information conflicts with other records in the database, for example, whether two obstacles were marked at the same location or within a similar timeframe. The updated secondary obstacle data can be rapidly fed back to the vehicle's autonomous driving system for real-time path planning and obstacle avoidance decisions. The updated data can also be properly stored for subsequent analysis and long-term retrieval. The storage system can have data archiving and retrieval capabilities, allowing users or maintenance personnel to query obstacle information at specific times and locations as needed for accident analysis, environmental monitoring or system performance evaluation.

[0050] The update process described above ensures the real-time and accuracy of obstacle data, providing a fresh and reliable source of information for autonomous driving. For autonomous vehicles in mining scenarios, this helps them adjust their driving strategies in a timely manner to avoid collisions or becoming trapped near obstacles. Regularly updating and improving the obstacle database creates a self-improving mechanism, accumulating more information about the environment and obstacles over time, gradually improving the overall performance and adaptability of obstacle detection.

[0051] In an embodiment of the present invention, first, in response to a work vehicle detecting first obstacle data, the first obstacle data may be matched with second obstacle data. Here, the first obstacle data is obstacle data identified based on the work vehicle's current perception data, and the second obstacle data is obstacle data identified based on the work vehicle's historical perception data at previous moments. Next, in response to a successful match between the first obstacle data and the second obstacle data, a fusion parameter is determined based on the obstacle type in the first obstacle data. Next, based on the fusion parameter, a first obstacle shape in the first obstacle data is fused with a second obstacle shape in the second obstacle data to obtain an initial obstacle shape. Next, the initial obstacle shape is smoothed to obtain a target obstacle shape. Finally, the second obstacle data is updated based on the target obstacle shape. It is easy to note that the present application utilizes the fusion of historical data and current data, and through smoothing, achieves a more accurate and stable description of obstacle shapes. Specifically, obstacle data detected in real time by the work vehicle is matched with data from historical moments. When the match is successful, the fusion parameters can be dynamically determined based on the obstacle type. The determination of the fusion parameters can take into account the physical characteristics of different obstacle types and their impact on the work vehicle's driving. By dynamically adjusting the fusion parameters, the characteristics of different obstacle types can be adapted. Based on the determined fusion parameters, the currently detected obstacle shape can be fused with the historical shapes to generate a preliminary fused initial obstacle shape. The initial obstacle shape is then smoothed to eliminate potential noise and improve the quality of the shape description. This can effectively offset errors that occur in single-moment detection and enhance the continuity and stability of the shape description. Based on the processed target obstacle shape, the historical obstacle data is updated, achieving iterative data improvement, further improving the accuracy of subsequent obstacle detection by the work vehicle, thereby resolving the technical problem of low stability and accuracy of obstacle detection by work vehicles in the related art.

[0052] In an embodiment of the present invention, the fusion parameters include a first projection parameter and a second projection parameter, where the first projection parameter represents a weight for fusing the first obstacle shape, and the second projection parameter represents a weight for fusing the second obstacle shape. Fusing the first obstacle shape in the first obstacle data with the second obstacle shape in the second obstacle data to obtain an initial obstacle shape includes: projecting, based on the first projection parameter, a plurality of first vertices constituting the first obstacle shape onto the second obstacle shape to obtain a plurality of first projection points; back-projecting, based on the second projection parameter, a plurality of second vertices constituting the second obstacle shape onto the first obstacle shape to obtain a plurality of second projection points; and generating the initial obstacle shape based on the plurality of first projection points and the plurality of second projection points.

[0053] The first projection parameter mentioned above may refer to the weight that affects the obstacle shape recognized at the current moment in the fusion result during the obstacle shape fusion process. A higher first projection parameter may indicate a greater tendency to trust the obstacle shape perceived in real time.

[0054] The second projection parameter mentioned above may refer to the weight of the obstacle shape affecting the historical moment in the fusion result. A lower second projection parameter may reflect that the historical data accounts for a smaller proportion in the fusion result.

[0055] The aforementioned multiple first vertices may refer to the shape of the obstacle identified at the current moment, such as a series of vertices on a polygon. The multiple first vertices provide geometric details of the obstacle outline.

[0056] The aforementioned multiple first projection points may refer to a series of new points formed by projecting multiple first vertices onto the obstacle shape at a historical moment. The multiple first projection points may reflect the position mapping of the first vertices on the second obstacle shape.

[0057] The aforementioned multiple second vertices may refer to obstacle shapes identified at historical moments, and the multiple second vertices may represent points on previously recorded obstacle contours.

[0058] The aforementioned multiple second projection points may refer to reversely projecting multiple second vertices onto the obstacle shape at the current moment to generate a series of new points as the other side of the fusion process.

[0059] In an optional embodiment, the fusion parameters may include a first projection parameter and a second projection parameter. The first projection parameter may represent the weight of the current data in the fusion process, while the second projection parameter may represent the weight of the historical data in the fusion process. For example, in a mining scenario, for small obstacles, the first projection parameter may be set to 0.9 to indicate the importance of the current data; the second projection parameter may be set to 0.1 to maintain the reference role of the historical data, but with a lower weight. Specifically, the first vertices that form the first obstacle shape can be projected onto the second obstacle shape based on the first projection parameter. The projection process may include: finding the edge in the second obstacle shape closest to the first vertex as the target edge; constructing a perpendicular line from the first vertex to the target edge; and calculating the intersection point of the perpendicular line with the target edge, which is the first projection point. This process can be repeated for each first vertex to generate a set containing all first projection points. Next, the second vertices that form the second obstacle shape can be back-projected onto the first obstacle shape based on the second projection parameter. This process includes finding the closest edge, constructing the perpendicular line, and calculating the intersection point, ultimately generating another set containing all the second projection points. After obtaining the first projection point and the second projection point set, a new, smooth and stable obstacle shape description, i.e., the initial obstacle shape, can be generated based on the multiple first projection points and the multiple second projection points and the weights of the projection parameters of the multiple first projection points and the multiple second projection points through interpolation, fitting, or polygon reconstruction.

[0060] In this process, weight distribution balances the immediacy of current perception data with the stability of historical data. Due to its high weight, current data dominates the shape description, ensuring rapid adaptation to environmental changes. Historical data serves as a supplement, helping to overcome transient environmental interference and maintain the coherence of the description. The fusion of obstacle shapes reduces misidentification or missed detections caused by fluctuations in data at a single moment. By integrating data from multiple moments, a more comprehensive and stable obstacle model can be constructed, maintaining high detection efficiency and accuracy even in dynamic environments such as mines.

[0061] In an embodiment of the present invention, based on first projection parameters, multiple first vertices used to constitute the first obstacle shape are projected onto the second obstacle shape to obtain multiple first projection points, including: generating multiple second edges used to constitute the second obstacle shape based on multiple second vertices used to constitute the second obstacle shape; determining a target edge among the multiple second edges based on the distance between the first vertex and the second edge, wherein the distance between the first vertex and the target edge is smaller than the distance between the first vertex and the other edges, and the other edges are any edges among the multiple second edges except the target edge; and projecting the first vertex onto the target edge based on the first projection parameters to obtain the first projection point.

[0062] The aforementioned multiple second edges may refer to edges on the polygonal outline of the obstacle shape identified at a historical moment. The multiple second edges are formed by connecting multiple second vertices, i.e., vertices of the obstacle outline at a historical moment, and together constitute a geometric boundary description of the obstacle.

[0063] The target edge may be an edge selected from the plurality of second edges that is closest to the first vertex based on the distances from the obstacle vertex to each edge at the current moment. During the fusion process, the first vertex may be projected onto the target edge based on the first projection parameters to generate a first projection point.

[0064] In an optional embodiment, multiple edges can first be generated based on the multiple vertices that comprise the second obstacle shape. This process can include determining the connectivity of each vertex within the obstacle shape, specifically connecting adjacent vertices to form a closed polygonal boundary. In computer graphics and geometry processing, this step can be implemented using a vertex list and an index array, where the index array indicates which vertices should be connected to form edges. Next, the distance from each first vertex of the first obstacle shape to each edge of the second obstacle shape can be calculated. This can be done using a point-to-line segment perpendicular distance calculation method. For each edge, the shorter distance from the first vertex to the edge can be found. The formula for point-to-line segment distance can be used to calculate the perpendicular distance through vector operations, taking into account the coordinates of the line segment's two endpoints and the vertex coordinates. For each first vertex, the closest edge in the second obstacle shape can be identified as the target edge. This selection ensures that the generated projection point is based on similar geometric features, thereby improving the accuracy and effectiveness of the projection algorithm. The target edge can be identified by comparing all calculated distances and selecting the edge with the smaller value. After the target edge is determined, the first projection parameter can be used to project the first vertex onto the target edge to generate the first projection point. The projection point can be calculated using linear algebra, which can be specifically as follows: determine the normal vector of the target edge, use the point-to-line distance formula combined with the first projection parameter, and calculate the position of the projection point on the target edge. This process ensures that the fusion of the current moment data and the historical moment data is based on the geometric characteristics of the shape. Through weight adjustment, it can balance the immediacy of current information and the stability of historical information.

[0065] In the above process, geometrically-based distance calculation and projection ensure the geometric consistency of the first and second obstacle shapes during fusion. This helps ensure the coherence of the obstacle shape description between successive moments, avoiding sudden changes or distortion. Selecting the edge closest to the first vertex as the target edge for projection effectively reduces errors during data fusion. A closer distance, meaning a higher similarity in geometric features, results in a projected point that more closely resembles the actual obstacle shape, thereby improving the accuracy of the fused shape.

[0066] In an embodiment of the present invention, based on the first projection parameter, the first vertex is projected onto the target edge to obtain the first projection point, including: determining the intersection of the first vertex and the target edge, wherein the straight line formed by the first vertex and the intersection is perpendicular to the target edge; generating a direction vector based on the coordinates of the first vertex and the coordinates of the intersection; and obtaining the coordinates of the first projection point based on the direction vector, the first projection parameter and the coordinates of the first vertex.

[0067] The intersection point can be the foot point of the perpendicular line between the first vertex and the target edge. In other words, if a straight line perpendicular to the target edge is drawn from the first vertex, the intersection point of the straight line and the target edge can be the intersection point.

[0068] The above-mentioned direction vector may refer to a vector indicating a direction determined by a line between the first vertex and the intersection point between the first vertex and the target edge, and may reflect a direction from the first vertex to the intersection point.

[0069] In an optional embodiment, the perpendicular intersection point from the first vertex to the target edge can be determined first. This process can include finding the intersection point of a line perpendicular to the target edge and the target edge starting from the first vertex. The specific steps can be as follows: determining the starting and ending point coordinates of the target edge, calculating the direction vector and normal vector of the target edge; constructing a line perpendicular to the target edge based on the first vertex, where the intersection point of the line and the target edge is the desired perpendicular intersection point. Calculating the intersection coordinates can include solving a system of linear equations, considering the mathematical expressions of the line and edge, and finding the coordinates of the intersection point of the first vertex and the target edge. After obtaining the intersection point of the first vertex and the target edge, a direction vector can be constructed from the first vertex to the intersection point. The direction vector can be constructed by using the intersection coordinates and the first vertex coordinates to calculate the difference between the two coordinates. The resulting direction vector indicates the direction and distance of the projection. Finally, the coordinates of the first projection point can be calculated based on the constructed direction vector, the first projection parameters, and the coordinates of the first vertex. The specific implementation method can be as follows: use the first projection parameter to adjust the length of the direction vector to control the distance between the projection point and the first vertex; add the adjusted direction vector to the coordinates of the first vertex to obtain the final coordinates of the first projection point. This process can ensure the accuracy of the projection operation. By adjusting the projection parameters and the length of the direction vector, the position of the projection point can be controlled so that the position of the projection point can reflect the optimal projection state from the first vertex to the target edge.

[0070] The precise generation of perpendicular intersection points and direction vectors in the above process significantly improves the geometric accuracy of obstacle shape fusion. By ensuring that the projection points fall precisely on a line perpendicular to the target edge, detailed variations in the obstacle shape can be captured more accurately, avoiding shape distortion caused by projection errors. The coordinate calculation of the first projection point, combined with the adjustment of the first projection parameters, provides a stable and accurate description of the obstacle shape for fusion.

[0071] In an embodiment of the present invention, the coordinates of the first projection point are obtained based on the direction vector, the first projection parameter and the coordinates of the first vertex, including: obtaining the product of the first projection parameter and the direction vector to obtain the coordinate offset; obtaining the sum of the coordinate offset and the coordinates of the first vertex to obtain the coordinates of the first projection point.

[0072] The coordinate offset described above refers to the geometric distance required to move from the first vertex to its projection point on the target edge, expressed as increments in the X and Y directions in the coordinate system. The coordinate offset is determined by multiplying the first projection parameter by the direction vector and reflects the specific implementation of the projection operation in spatial coordinates.

[0073] In an optional embodiment, a coordinate offset can be first calculated. This can be obtained by multiplying the first projection parameter by the product of the direction vector. This process may include performing a vector scaling operation. The specific steps may be as follows: obtain the X and Y components of the direction vector; multiply the X and Y components of the direction vector by the first projection parameter to obtain a scaled direction vector, i.e., the coordinate offset. This multiplication operation scales the direction vector proportionally along its direction, thereby controlling the distance of the projected point from the first vertex. After obtaining the coordinate offset, the coordinate offset can be added to the coordinates of the first vertex to obtain the coordinates of the first projected point. This calculation step can be decomposed into the following operations: add the X coordinate of the first vertex to the X component of the coordinate offset to obtain the X coordinate of the first projected point. Add the Y coordinate of the first vertex to the Y component of the coordinate offset to obtain the Y coordinate of the first projected point. Through this calculation, the first projected point is placed in the direction indicated by the direction vector, with a certain displacement relative to the first vertex, which is controlled by the first projection parameter.

[0074] In the above process, by calculating the coordinate offset by multiplying the first projection parameter and the direction vector, the position of the first projection point relative to the first vertex can be accurately controlled. The size of the first projection parameter reflects the weight of the data at the current moment in the fusion process. A larger first projection parameter value means that the current data is more important, and the projection point tends to be closer to the position indicated by the current data.

[0075] In an embodiment of the present invention, smoothing an initial obstacle shape to obtain a target obstacle shape includes: determining the number of initial vertices used to constitute the initial obstacle shape; in response to the number of initial vertices being less than or equal to a first preset number, determining the initial obstacle shape to be the target obstacle shape; and in response to the number of initial vertices being greater than the first preset number, filtering the initial vertices to obtain filtered vertices, and generating the target obstacle shape based on the filtered vertices.

[0076] The above-mentioned initial vertices may refer to the vertices used to constitute the initial obstacle shape after the obstacle shape fusion processing is completed, that is, the vertices of the fused obstacle polygon outline. The initial vertices may include the points projected from the perception data at the current moment and the historical moment.

[0077] The above-mentioned first preset number can refer to a pre-set numerical threshold for judging whether the number of vertices of the initial obstacle shape is too large, thereby deciding whether vertex filtering is required. The first preset number can be 5, and can also be determined according to actual needs. It is not limited here. The selection of the first preset number is intended to maintain the simplicity and accuracy of the shape description and avoid redundant information and additional computational burden caused by too many vertices.

[0078] The filtered vertices mentioned above can refer to the representative vertices that are crucial to describing the obstacle shape, which are retained after being screened and improved by a specific algorithm based on the initial vertices. This helps to remove redundant or overly subtle details, making the description of the obstacle shape more concise and stable, while also reducing the demand for computing resources.

[0079] In an optional embodiment, the number of vertices that comprise the initial obstacle shape can first be counted. This counting process can be used to assess whether further vertex filtering is necessary. Next, based on the counted initial vertex count, the initial vertex count can be compared with a first preset number. The first preset number can be set based on a trade-off between the complexity of the obstacle shape description and computational efficiency, with different thresholds applicable to different application environments. If the initial vertex count is less than or equal to the first preset number, the obstacle shape can be determined to be sufficiently smooth and concise, eliminating the need for further vertex filtering. The initial obstacle shape can then be directly used as the target obstacle shape for subsequent autonomous driving decisions. When the initial vertex count exceeds the first preset number, a vertex filtering mechanism can be initiated. This process can include identifying and retaining critical boundary vertices that define the obstacle shape's outline, filtering out excessive internal vertices, such as those that contribute little to the overall shape description, and fitting or reconstructing the filtered vertex set to generate a simpler, smoother target obstacle shape.

[0080] In the above process, by controlling the number of vertices, the complexity of the obstacle shape description can be improved. For specific environments such as mining scenes, too many vertices will increase the computational burden and cause the shape description to be too detailed, affecting the overall smoothness and stability. Appropriately reducing the number of vertices can help simplify the obstacle shape and improve the efficiency of the algorithm. The vertex filtering mechanism balances the relationship between the accuracy of the obstacle shape description and computational efficiency. On the premise of ensuring that the shape description is not distorted due to simplification, reducing the number of vertices can significantly reduce the computational requirements of subsequent processing steps and improve the overall response speed of the algorithm. In a mining environment, obstacles will cause slight fluctuations in shape due to factors such as vehicle vibration and environmental changes. Through vertex filtering and shape reconstruction, a more stable and consistent obstacle shape description can be generated.

[0081] In an embodiment of the present invention, initial vertices are filtered to obtain filtered vertices, including: determining multiple boundary points in the initial vertices based on the coordinates of the initial vertices, wherein different boundary points are used to represent the boundaries of the initial obstacle shape in different directions; determining an initial vertex located between two adjacent boundary points to obtain at least one target pinch point, wherein two adjacent boundary points are used to represent the boundaries of the initial obstacle shape in two adjacent directions; filtering the at least one target pinch point to obtain filtered pinch points, wherein the number of filtered pinch points is less than or equal to a second preset number; and obtaining filtered vertices based on the filtered pinch points and the multiple boundary points.

[0082] The aforementioned multiple boundary points can refer to vertices located in different directions on the obstacle's outline, and can be used to identify the obstacle's outer boundaries or turning points in these directions. In the polygon after shape fusion and smoothing, these multiple boundary points can help determine the shape and size of the entire obstacle.

[0083] The target clamp point mentioned above may refer to an initial vertex between multiple boundary points. The target clamp point is sandwiched between the boundary points and may be used to refine the description of the obstacle shape.

[0084] The aforementioned filtered pinch points may refer to pinch points that are retained after removing redundant points from the target pinch points and are considered sufficiently critical for shape description. By determining the filtered pinch points, a more concise and accurate obstacle shape description can be generated.

[0085] The above-mentioned second preset number can refer to a pre-set threshold for limiting the number of filter pinch points. The first preset number can be 2, and can also be determined according to actual needs. It is not limited here. By controlling the number of filter pinch points below the second preset number, it can be ensured that the shape of the obstacle is neither too simplified nor too complicated, achieving a better description balance point.

[0086] In an optional embodiment, key vertices representing the obstacle boundary in different directions can be selected from the initial set of vertices that constitute the obstacle shape based on vertex coordinates as boundary points. Boundary points can be determined based on the maximum and minimum values ​​of the vertices on the X and Y coordinate axes, as well as their position within the obstacle shape, to ensure a complete description of the obstacle's outline. Next, vertices located between adjacent boundary points, known as target pinch points, can be identified. The identification of target pinch points can include analyzing the continuity of the obstacle shape boundary and the distribution of vertices, with the goal of identifying excessive internal vertices that contribute little to the shape description. The identified target pinch points can then be filtered to retain key vertices and remove redundant vertices. The filtering mechanism can be based on specific filtering rules, such as the distance between the point and the edge, the angular difference between the points, etc., to ensure that the number of filtered vertices is within a second preset number while maintaining the integrity and smoothness of the obstacle shape. Finally, a filtered vertex set can be generated based on the filtered grip points and retained boundary points to construct a smooth and concise target obstacle shape. This process can include steps such as geometric reconstruction and polygon fitting to ensure that the shape description reflects the key features of the obstacle while avoiding excessive details.

[0087] In the above process, filtering of boundary points and target pinch points simplifies the obstacle shape description, removes redundant vertices, and generates a simpler and smoother target shape, which helps improve the efficiency of shape description, reduces computing resource consumption, and enhances shape stability. The target pinch point filtering mechanism can enhance the robustness of obstacle detection and adapt to different types of obstacles and changing environmental conditions. In mining scenarios, the shape of obstacles can be affected by various factors, such as vehicle vibration and terrain changes. By intelligently filtering vertices, a more stable and consistent shape description can be generated, reducing misidentification or emergency obstacle avoidance caused by shape fluctuations.

[0088] In an embodiment of the present invention, filtering at least one target pinch point to obtain a filtered pinch point includes: in response to at least one target pinch point including one target pinch point, determining the target pinch point as a filtered pinch point; in response to at least one target pinch point including multiple target pinch points, filtering the multiple target pinch points to obtain a filtered pinch point.

[0089] In an optional embodiment, when at least one identified target grip actually only contains one grip, the target grip can be directly determined as a filtered grip. When at least one target grip contains multiple grips, these grips are further analyzed and filtered to determine which should ultimately be retained as filtered grips and which should be removed. This process can be based on the following methods: vertices farther from the boundary point can be retained, as these vertices are more likely to reflect the true characteristics of the obstacle shape, while vertices closer to the boundary point may be the result of noise or excessive detail. Alternatively, vertices with larger angles that change the direction of the obstacle shape can be retained, while vertices with smaller angles of change can be removed to maintain the turning points and edge features of the shape while removing unnecessary details. Alternatively, vertices with longer projection lines can be retained, as these vertices play a more important role in the shape description, while vertices with shorter projection lines may be considered redundant and thus filtered out. The specific method for filtering multiple target grips can be determined based on actual needs and is not limited here.

[0090] In the above process, the target grip filtering mechanism effectively improves the sophistication of obstacle shape descriptions. By distinguishing between single and multiple target grips, the complexity of the shape description can be intelligently controlled. Through intelligent selection and filtering, the essential features of the shape are preserved while overly refined areas are removed, helping to balance algorithm efficiency and shape description accuracy. By setting clear filtering criteria, it is possible to remove redundant information while retaining vertices that are critical to the shape description, ensuring that shape smoothing is fast and distortion-free.

[0091] In an embodiment of the present invention, a plurality of target pinch points are filtered to obtain filtered pinch points, including: constructing a boundary line based on the coordinates of two adjacent boundary points; determining a projection line of any target pinch point on the boundary line, wherein the projection line passes through the target pinch point and is perpendicular to the boundary line; and filtering a plurality of target pinch points based on the length of the projection line to obtain filtered pinch points, wherein the length of the projection line corresponding to the filtered pinch point is greater than the length of the projection line corresponding to other pinch points, and the other pinch points are any target pinch points other than the filtered pinch point among the plurality of target pinch points.

[0092] The boundary line described above can be a line segment constructed by the coordinates of two adjacent boundary points. The boundary line can be used to define the outer contour or boundary of the obstacle shape in a specific direction. In polygonal shape descriptions, the boundary line can be regarded as the edge connecting the polygon's vertices, helping to define the shape's scope.

[0093] The projection line, a line segment perpendicular to the boundary line and originating from the target grip, can be used to measure the shortest distance from the target grip to the boundary line. Projection lines provide distance information and help understand the position of the target grip relative to the boundary line, aiding in shape refinement and streamlining.

[0094] In an optional embodiment, a boundary line can first be constructed based on the coordinates of two adjacent boundary points. The boundary line can be used to define the outline of the obstacle shape in a specific direction and can serve as the basis for subsequent projection line length calculation. For each target clamp point, the projection line of each target clamp point on the boundary line can be calculated. The calculation of the projection line can be based on the perpendicular relationship between the target clamp point and the boundary line to ensure that the projection line accurately passes through the target clamp point and is perpendicular to the boundary line. This process can include the point-to-line distance calculation in linear algebra, and the projection line is obtained by solving the shorter distance from the target clamp point to the boundary line. Then, the projection line lengths corresponding to each target clamp point can be compared, and the clamp points can be filtered based on these length values. Specifically, the target clamp point with a longer projection line length can be selected as the filter clamp point, and the clamp point with a shorter projection line length can be removed. This filtering mechanism can ensure the retention of key feature points, while removing those vertices that have little impact on the shape description, so as to simplify the obstacle shape and improve the smoothness of the shape description.

[0095] In the above process, the filtering mechanism based on the projection line length can effectively identify and retain the vertices that are more important in the shape description, while removing redundant or detailed grip points, which helps to generate a more concise and easier to handle obstacle shape description and reduces the complexity of subsequent calculation steps.

[0096] In an embodiment of the present invention, fusing a first obstacle shape in the first obstacle data with a second obstacle shape in the second obstacle data to obtain an initial obstacle shape includes: determining the number of first vertices constituting the first obstacle shape; and, in response to the number of first vertices being greater than or equal to a third preset number, fusing the first obstacle shape with the second obstacle shape to obtain the initial obstacle shape.

[0097] The third preset number mentioned above may refer to a pre-set numerical threshold for determining whether the number of vertices constituting the first obstacle shape reaches or exceeds the appropriate level required for fusion processing. The third preset number may be 3, or may be determined based on actual needs. This is not limited here. The third preset number may be set based on specific obstacle detection and fusion requirements to ensure that sufficient details are taken into account during the shape fusion process while avoiding computational efficiency issues caused by an excessive number of vertices.

[0098] In an optional embodiment, the number of vertices comprising the first obstacle shape detected at the current moment can be counted. This counting step can be used to subsequently determine the number of vertices and select a fusion strategy. Next, the first number of vertices can be compared with a third preset number, which can be set based on empirical data, to determine whether the current obstacle shape description is too complex and can be simplified and smoothed through fusion. If the number of first vertices is greater than or equal to the third preset number, it can be determined that the current shape description is too detailed and contains redundant information. In this case, fusion processing can be initiated to merge the current shape with the shape at the previous moment to generate a simpler and smoother initial obstacle shape.

[0099] The above process, by determining the number of vertices, intelligently identifies when fusion processing is needed to improve the shape description. When the number of vertices exceeds a third preset number, fusion processing effectively reduces the number of vertices, simplifying the shape description and avoiding the waste of computing resources and processing delays caused by overly complex shapes. By fusing the current shape with historical shapes, shape changes are smoothed, reducing shape fluctuations, ensuring a more stable obstacle shape description, and avoiding sudden braking or path replanning caused by sudden shape changes in the operating vehicle.

[0100] In an embodiment of the present invention, determining the fusion parameters based on the obstacle type in the first obstacle data includes: in response to the obstacle type being a preset type, determining that the fusion parameters include the first projection parameter and the second projection parameter.

[0101] The above-mentioned preset types can refer to obstacle categories or attributes that are suitable for specific fusion strategies, such as small obstacles that require special attention to shape stability, such as small stones or ruts in mining scenes. The shape and position of the preset type can change with time and environmental factors, but the overall outline is relatively fixed.

[0102] In an optional embodiment, the detected obstacle type can first be determined. This determination can be based on obstacle characteristics such as size, shape, and location to determine whether the obstacle meets preset small obstacle classification criteria. This process can include machine learning model recognition or rule-based feature comparison to ensure that fusion parameter adjustments are tailored to the specific obstacle type. Once the obstacle is confirmed to be a preset type, such as a small obstacle in a mining scenario, fusion parameters can be automatically determined, including first and second projection parameters. The first and second projection parameters can influence the weighting of shape fusion and smoothing, as well as the degree of mutual influence between the first and second obstacle shapes. The first projection parameters can be used to adjust the weight of the currently detected obstacle shape in the fusion process, while the second projection parameters can adjust the weight of obstacle shapes at previous times. For example, if the first obstacle shape is deemed more accurate or reliable, the first projection parameters can be set to a higher value. After the first and second projection parameters are determined, shape fusion and smoothing can be performed. The fusion process may include performing operations such as projection and translation on vertices of the first obstacle shape and the second obstacle shape to generate an initial obstacle shape that combines the features of the first obstacle shape and the second obstacle shape.

[0103] By dynamically adjusting the fusion parameters in this process, the shape fusion process can be more precisely controlled, ensuring that the fused shape reflects the latest perception data while also taking into account the continuity of historical data. This helps improve the accuracy of obstacle recognition in autonomous vehicles. The fusion parameter settings for pre-set obstacle types allow for flexible adjustment of fusion and smoothing strategies based on different obstacle types and operational scenarios. For example, in a mining scenario, the shapes of small obstacles can be smoothed to reduce shape fluctuations caused by external factors such as vehicle vibration.

[0104] In an embodiment of the present invention, the method further includes: in response to a failure in matching the first obstacle data with the second obstacle data, merging the first obstacle data with the second obstacle data.

[0105] In an optional embodiment, during the obstacle detection and identification process, current perception data can be matched with perception data from historical moments to continuously monitor changes in the status of the same obstacle and provide continuous and consistent obstacle information. If the first and second obstacle data cannot be successfully matched, this may be due to factors including, but not limited to, differences in obstacle location, size, shape, etc. exceeding a preset threshold, or the inability to find an obstacle instance matching the current data in the historical data. If matching fails, the first and second obstacle data can be merged. This merging can include combining the obstacle information from the first and second obstacle data to form a new, more comprehensive description of the obstacle. Data merging can be performed using a variety of methods, such as averaging, weighted averaging, or prioritizing newer data to selectively merge key information from the two datasets to generate updated obstacle data.

[0106] In the above process, if the first obstacle data fails to match the second obstacle data, the first obstacle data and the second obstacle data are merged to update the second obstacle data. By integrating current and historical data, more comprehensive historical obstacle information can be obtained, enhancing the adaptability of obstacle detection and better coping with the ever-changing working environment. In mining scenarios, the position and shape of obstacles will change rapidly due to the movement of working vehicles and changes in terrain. The merging strategy can help to timely update obstacle information and improve the real-time and accuracy of obstacle detection.

[0107] The following describes the technical solution proposed in this application in conjunction with an optional embodiment. This application proposes a method for smoothing the shape of small obstacles in mining scenarios. Mining trucks often encounter small obstacles such as rocks and high ruts when operating in mining scenarios. Focusing on mining scenarios, this application proposes a novel small obstacle shape smoothing fusion method that can stably output the shape of small obstacles, improving the efficiency of autonomous driving operations.

[0108] The method for smoothing the shape of small obstacles in mining scenarios proposed in this application may include the following steps: obtaining a perception obstacle queue, which may be first obstacle data; matching elements in the perception obstacle queue with obstacle elements in a tracking queue, which may be second obstacle data. If the match fails, a target element may be added to the tracking queue. The target element may refer to obstacle data identified by the perception data at the current moment, i.e., the first obstacle data. In other words, in response to a failure to match the first and second obstacle data, the first and second obstacle data are merged. If the match succeeds, a determination may be made as to whether the target element is a small obstacle. If not, the corresponding tracking queue element parameters are directly tracked. If the target element is a small obstacle type, the small obstacle shape is smoothed. In other words, in response to the obstacle type being a preset type, the small obstacle smoothing process is performed.

[0109] Figure 2 is a schematic diagram of an optional obstacle detection process according to an embodiment of the present invention, such as Figure 2 As shown, in response to the work vehicle detecting first obstacle data, the first obstacle data is matched with the second obstacle data. A determination is made as to whether the first obstacle data and the second obstacle data are successfully matched. If not, the first obstacle data and the second obstacle data are merged. If so, a determination is made as to whether the obstacle type in the first obstacle data is a preset type. If not, the second obstacle data is updated based on the target obstacle shape. If so, the first obstacle shape in the first obstacle data is merged with the second obstacle shape in the second obstacle data to obtain an initial obstacle shape, the initial obstacle shape is smoothed to obtain a target obstacle shape, and the second obstacle data is updated based on the target obstacle shape. A jump loop is executed, and in response to the work vehicle detecting the first obstacle data, the first obstacle data and the second obstacle data are matched.

[0110] Figure 3 is a schematic diagram of an optional obstacle smoothing process according to an embodiment of the present invention, such as Figure 3 As shown, the number of first vertices used to form a first obstacle shape is determined. A determination is made as to whether the number of first vertices is greater than or equal to a third preset number. If not, the obstacle shape smoothing process is not performed. If so, the first obstacle shape is merged with the second obstacle shape to obtain an initial obstacle shape. A determination is made as to whether the number of initial vertices used to form the initial obstacle shape is greater than the first preset number. If not, a target obstacle shape is generated. If so, the initial vertices used to form the initial obstacle shape are filtered to generate the target obstacle shape.

[0111] In this application, the smoothing of small obstacle shapes may include the following steps: Input parameters may be verified, such as if the number of polygon points is greater than or equal to 3, that is, in response to the number of first vertices being greater than or equal to a third preset number, the small obstacle is smoothed; if the verification fails, no subsequent processing is performed. Next, the polygons of the small obstacle may be fused and smoothed, which may include projecting the points of the small obstacle polygon from the sensing queue onto the small obstacle polygon corresponding to the tracking queue based on a first projection parameter, such as 0.9 (not limited here), to generate new points. Next, the points of the small obstacle polygon (polyon) of the tracking source can be projected onto the small obstacle polygon (polyon) corresponding to the perception queue based on the second projection parameter, such as 0.1, which is not limited here, to generate new points. Based on the newly generated points, a new polygon (polyon) is recalculated. That is, based on the first projection parameter, multiple first vertices used to constitute the first obstacle shape are projected onto the second obstacle shape to obtain multiple first projection points; based on the second projection parameter, multiple second vertices used to constitute the second obstacle shape are reversely projected onto the first obstacle shape to obtain multiple second projection points. Based on the multiple first projection points and the multiple second projection points, an initial obstacle shape is generated.

[0112] Figure 4 FIG. 1 is a schematic diagram of an initial obstacle shape obtained by an optional fusion process according to an embodiment of the present invention, such as Figure 4 As shown, based on the first projection parameters, multiple first vertices used to constitute the first obstacle shape are projected onto the second obstacle shape to obtain multiple first projection points; based on the second projection parameters, multiple second vertices used to constitute the second obstacle shape are reversely projected onto the first obstacle shape to obtain multiple second projection points; based on the multiple first projection points and the multiple second projection points, an initial obstacle shape is generated.

[0113] Next, a new polygon can be generated. A determination is made as to whether the number of points in the fused and smoothed polygon is greater than five, i.e., a first preset number (not limited here). If the number is less than or equal to the first preset number, the fused and smoothed polygon can be directly output. If the number is greater than the first preset number, the number of points in the polygon can be adjusted. Specifically, the number of initial vertices used to form the initial obstacle shape is determined. If the number of initial vertices is less than or equal to the first preset number, the initial obstacle shape is determined to be the target obstacle shape. If the number of initial vertices is greater than the first preset number, the initial vertices are filtered to obtain filtered vertices, and the target obstacle shape is generated based on the filtered vertices. Finally, the fused and smoothed polygon, i.e., the target obstacle shape, can be output.

[0114] The above-mentioned point of the small obstacle polygon (polyon) of the sensing queue source is projected onto the small obstacle polygon (polyon) corresponding to the tracking queue based on the first projection parameter according to the first projection parameter, which is not limited here, to generate a new point. The following steps can be included: calculating the distance from the point to each polygon (polyon) edge, finding the polygon (polyon) edge closer to the point, that is, based on the distance between the first vertex and the second edge, determining the target edge among the multiple second edges; then, the intersection (intersection_point) of the point (point) and the closer polygon (polyon) edge can be calculated, that is, determining the intersection (intersection_point) of the first vertex and the target edge; then, the direction vector (direction) between the point (point) and the intersection (intersection_point) can be calculated, that is, based on the coordinates of the first vertex and the coordinates of the intersection, a direction vector is generated. The calculation formula can be expressed as follows:

[0115] direction.x = intersection.x - point.x;

[0116] direction.y = intersection.y - point.y;

[0117] Among them, the above-mentioned direction.x can represent the value of the horizontal coordinate (X) of the direction vector (direction), the above-mentioned direction.y can represent the value of the vertical coordinate (Y) of the direction vector (direction), the above-mentioned intersection.x can represent the value of the horizontal coordinate (X) of the intersection point (intersection_point), the above-mentioned intersection.y can represent the value of the vertical coordinate (Y) of the intersection point (intersection_point), the above-mentioned point.x can represent the point (point), that is, the value of the horizontal coordinate (X) of the first vertex, and the above-mentioned point.y can represent the point (point), that is, the value of the vertical coordinate (Y) of the first vertex.

[0118] Next, the newly generated point of the projection can be calculated based on the direction vector (direction) and the projection parameter (param). That is, based on the direction vector, the first projection parameter and the coordinates of the first vertex, the coordinates of the first projection point are obtained. The calculation formula of the coordinates of the first projection point can be expressed as follows:

[0119] new_point.x = point.x + param * direction.x;

[0120] new_point.y = point.y + param * direction.y;

[0121] The new_point.x mentioned above may represent the value of the horizontal coordinate (X) of the first projection point, the new_point.y mentioned above may represent the value of the vertical coordinate (Y) of the first projection point, and the param mentioned above may represent the first projection parameter.

[0122] Figure 5 is a schematic diagram of an optional process of obtaining multiple first projection points according to an embodiment of the present invention, such as Figure 5 As shown, based on multiple second vertices used to constitute the shape of the second obstacle, multiple second edges used to constitute the shape of the second obstacle are generated, and based on the distance between the first vertex and the second edge, a target edge among the multiple second edges is determined; the intersection of the first vertex and the target edge is determined; based on the coordinates of the first vertex and the coordinates of the intersection, a direction vector is generated; the product of the first projection parameter and the direction vector is obtained to obtain the coordinate offset; the sum of the coordinate offset and the coordinates of the first vertex is obtained to obtain the coordinates of the first projection point.

[0123] Adjusting the number of points in the polygon may include the following steps: converting the coordinates of the polygon to the vehicle coordinate system; then finding four boundary points: an upper boundary point, a lower boundary point, a left boundary point, and a right boundary point; then determining whether the number of points sandwiched between two boundary points is greater than a second preset number, which may be 1 and is not limited to this. If the number of points sandwiched between the two boundary points is less than or equal to the second preset number, outputting the polygon with the adjusted number of points; if the number of points sandwiched between the two boundary points is greater than the second preset number, calculating the length of the projection line of each pinch point, first removing points with short projection lines so that the number of points between the two boundary points is the second preset number, and finally outputting the polygon with the adjusted number of points. Specifically, based on the coordinates of an initial vertex, determining multiple boundary points within the initial vertex, determining an initial vertex located between two adjacent boundary points, obtaining at least one target pinch point, filtering the at least one target pinch point to obtain a filtered pinch point, and obtaining a filtered vertex based on the filtered pinch point and the multiple boundary points.

[0124] Figure 6 is a schematic diagram of an optional process of obtaining filtered vertices according to an embodiment of the present invention, such as Figure 6 As shown, based on the coordinates of the initial vertex, multiple boundary points in the initial vertex are determined. The initial vertex located between two adjacent boundary points is determined to obtain at least one target pinch point. It is determined whether the number of target pinch points is greater than a second preset number. If not, the target pinch point is determined to be a filtered pinch point, and a filtered pinch point is obtained. If so, a boundary line is constructed based on the coordinates of the two adjacent boundary points; a projection line of any target pinch point on the boundary line is determined; and multiple target pinch points are filtered based on the length of the projection line to obtain a filtered pinch point.

[0125] Figure 7 is a schematic diagram of an optional boundary line and projection line according to an embodiment of the present invention, such as Figure 7 As shown in the figure, multiple boundary points such as the upper boundary point A, the lower boundary point B, the left boundary point C and the right boundary point D, as well as multiple boundary lines constructed based on the multiple boundary points, can then be determined. The projection line a and the projection line b are shown in the figure.

[0126] According to another aspect of an embodiment of the present invention, an unmanned vehicle is further provided, which can execute the obstacle detection method of the above embodiment. The specific implementation method and preferred application scenario are the same as those of the above embodiment and will not be repeated here.

[0127] Figure 8 Schematic diagram of a functional module of an unmanned vehicle according to an embodiment of the present invention. Figure 8As shown, the vehicle includes the following: a matching module 802 , a determination module 804 , a fusion module 806 , a smoothing module 808 and an update module 810 .

[0128] The matching module 802 is configured to match the first obstacle data with the second obstacle data in response to the work vehicle detecting the first obstacle data, wherein the first obstacle data is obstacle data identified based on the work vehicle's current perception data, and the second obstacle data is obstacle data of obstacles identified based on the work vehicle's historical perception data at previous moments. The determination module 804 is configured to determine a fusion parameter based on the obstacle type in the first obstacle data in response to a successful match between the first obstacle data and the second obstacle data. The fusion module 806 is configured to fuse and smooth the first obstacle shape in the first obstacle data with the second obstacle shape in the second obstacle data based on the fusion parameter to obtain an initial obstacle shape. The smoothing module 808 is configured to smooth the initial obstacle shape to obtain a target obstacle shape. The updating module 810 is configured to update the second obstacle data based on the target obstacle shape.

[0129] The fusion parameters include a first projection parameter and a second projection parameter, the first projection parameter is used to represent the weight of fusing the first obstacle shape, and the second projection parameter is used to represent the weight of fusing the second obstacle shape; the fusion module is further used to project multiple first vertices used to constitute the first obstacle shape onto the second obstacle shape based on the first projection parameter to obtain multiple first projection points; based on the second projection parameter, reversely project multiple second vertices used to constitute the second obstacle shape onto the first obstacle shape to obtain multiple second projection points; and generate an initial obstacle shape based on the multiple first projection points and the multiple second projection points.

[0130] Among them, the fusion module is also used to generate multiple second edges used to constitute the second obstacle shape based on multiple second vertices used to constitute the second obstacle shape; based on the distance between the first vertex and the second edge, determine the target edge among the multiple second edges, wherein the distance between the first vertex and the target edge is smaller than the distance between the first vertex and the other edges, and the other edges are any edges among the multiple second edges except the target edge; based on the first projection parameter, project the first vertex to the target edge to obtain a first projection point.

[0131] Among them, the fusion module is also used to determine the intersection of the first vertex and the target edge, wherein the straight line formed by the first vertex and the intersection is perpendicular to the target edge; based on the coordinates of the first vertex and the coordinates of the intersection, a direction vector is generated; based on the direction vector, the first projection parameter and the coordinates of the first vertex, the coordinates of the first projection point are obtained.

[0132] The fusion module is further used to obtain the product of the first projection parameter and the direction vector to obtain a coordinate offset; and obtain the sum of the coordinate offset and the coordinates of the first vertex to obtain the coordinates of the first projection point.

[0133] The smoothing module is further configured to determine the number of initial vertices used to form an initial obstacle shape; in response to the number of initial vertices being less than or equal to a first preset number, determine the initial obstacle shape as a target obstacle shape; and in response to the number of initial vertices being greater than the first preset number, filter the initial vertices to obtain filtered vertices, and generate the target obstacle shape based on the filtered vertices.

[0134] Among them, the smoothing module is also used to determine multiple boundary points in the initial vertex based on the coordinates of the initial vertex, wherein different boundary points are used to characterize the boundaries of the initial obstacle shape in different directions; determine the initial vertex located between two adjacent boundary points to obtain at least one target pinch point, wherein two adjacent boundary points are used to characterize the boundaries of the initial obstacle shape in two adjacent directions; filter the at least one target pinch point to obtain a filtered pinch point, wherein the number of the filtered pinch points is less than or equal to a second preset number; and obtain a filtered vertex based on the filtered pinch point and the multiple boundary points.

[0135] The smoothing module is further configured to determine the target pinch point as a filtered pinch point in response to at least one target pinch point including one target pinch point; and to filter the multiple target pinch points to obtain a filtered pinch point in response to at least one target pinch point including multiple target pinch points.

[0136] Among them, the smoothing module is also used to construct a boundary line based on the coordinates of two adjacent boundary points; determine the projection line of any target pinch point on the boundary line, wherein the projection line passes through the target pinch point and is perpendicular to the boundary line; filter multiple target pinch points based on the length of the projection line to obtain filtered pinch points, wherein the length of the projection line corresponding to the filtered pinch point is greater than the length of the projection line corresponding to other pinch points, and the other pinch points are any target pinch points among the multiple target pinch points except the filtered pinch point.

[0137] The fusion module is further configured to determine the number of first vertices used to form a first obstacle shape; in response to the number of first vertices being greater than or equal to a third preset number, the first obstacle shape is fused with the second obstacle shape to obtain an initial obstacle shape.

[0138] The determination module is further configured to determine, in response to the obstacle type being a preset type, that the fusion parameters include the first projection parameter and the second projection parameter.

[0139] The determining module is further configured to merge the first obstacle data with the second obstacle data in response to a failure in matching the first obstacle data with the second obstacle data.

[0140] An embodiment of the present application further provides an electronic device, comprising: a communication unit for communicating with a target vehicle; a memory storing an executable program; and a processor for running the program, wherein the method of each embodiment of the present invention is executed when the program is running.

[0141] The above-mentioned communication unit can be a basic module or component for transmitting, receiving and processing information, and can be a hardware module, a software module, or a combination of a hardware module and a software module, used to ensure that data is effectively and reliably transmitted between the platform or vehicle executing the path planning method proposed in this application and the target vehicle. For example, when the above-mentioned path planning method proposed in this application is executed by a platform, the platform may include the above-mentioned electronic device, and then the platform can communicate with the target vehicle based on the above-mentioned communication unit, and send the generated first driving path and second driving path to the target vehicle, etc.; for example, when the above-mentioned path planning method proposed in this application is executed by a vehicle, the vehicle may include the above-mentioned electronic device, and then the vehicle can communicate with the target vehicle based on the above-mentioned communication unit, and send the generated first driving path and second driving path to the target vehicle, etc.; the above-mentioned communication unit may include modules such as a transmitter, a receiver, a modulator, a demodulator, an encoder, and a decoder. The specific structure of the communication unit can be determined according to actual needs and is not limited here.

[0142] The above-mentioned memory may refer to a device inside a computer for storing data and programs, and may include memory, hard disk, etc., wherein the memory may be used to temporarily store running programs and data, the hard disk may be used to store programs and data for a long time, and the memory may be used to enable the computer to read and write data, as well as execute programs; the above-mentioned processor may be responsible for executing instructions in computer programs and performing data processing, and may be responsible for controlling and executing various operations, including arithmetic operations, logical operations, data transmission, etc.

[0143] An embodiment of the present application further provides a computer-readable storage medium, which includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0144] The above-mentioned computer storage medium may refer to a medium in a computer memory used to store certain discontinuous physical quantities. Computer storage media mainly include semiconductors, magnetic cores, magnetic drums, magnetic tapes, laser disks, etc. The stored program included in the computer-readable storage medium may be a set of instructions that can be recognized and executed by a computer, running on an electronic computer, and serving as an information tool to meet certain needs of people.

[0145] An embodiment of the present application further provides a computer program product, including a computer program, which implements the methods in various embodiments of the present invention when executed by a processor.

[0146] The above-mentioned computer program product may refer to a software program that has been written, tested and released, which can be run on a computer or other device. The computer program product may include an application, an operating system, tool software, etc., which is used to implement specific functions or solve specific problems.

[0147] An embodiment of the present application further provides a computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0148] The above-mentioned non-volatile computer-readable storage medium may refer to a medium for storing data. The non-volatile computer-readable storage medium can keep the data from being lost when the power is off, and can be used to store long-term data, such as operating systems, applications and user files. The non-volatile storage medium may include hard disk drives, solid-state drives, optical disks and flash memory storage devices, etc.

[0149] The embodiments of the present application further provide a computer program, which implements the methods in the above-mentioned embodiments of the present invention when executed by a processor.

[0150] The above-mentioned computer program may refer to a collection of instructions used to tell a computer to perform a specific task or operation. A computer program may be written by a programmer using a specific programming language and may include algorithms, data structures, logic, and control flows. Computer programs may be used for a variety of purposes, including application software, operating systems, and the like.

[0151] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0152] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0153] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0154] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0155] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), a mobile hard drive, a magnetic disk, or an optical disk.

[0156] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. An obstacle detection method, characterized in that: include: In response to the work vehicle detecting first obstacle data, matching the first obstacle data with second obstacle data, wherein the first obstacle data is obstacle data identified based on perception data of the work vehicle at a current moment, and the second obstacle data is obstacle data identified based on historical perception data of the work vehicle at a historical moment; In response to a successful matching between the first obstacle data and the second obstacle data, determining a fusion parameter based on an obstacle type in the first obstacle data; fusing the first obstacle shape in the first obstacle data and the second obstacle shape in the second obstacle data based on the fusion parameter to obtain an initial obstacle shape; Smoothing the initial obstacle shape to obtain a target obstacle shape; updating the second obstacle data based on the target obstacle shape; The fusion parameters include a first projection parameter and a second projection parameter, the first projection parameter being used to represent a weight for fusing the first obstacle shape, and the second projection parameter being used to represent a weight for fusing the second obstacle shape. Fusing the first obstacle shape in the first obstacle data with the second obstacle shape in the second obstacle data to obtain an initial obstacle shape includes: projecting a plurality of first vertices constituting the first obstacle shape onto the second obstacle shape based on the first projection parameters to obtain a plurality of first projection points; back-projecting a plurality of second vertices constituting the second obstacle shape onto the first obstacle shape based on the second projection parameters to obtain a plurality of second projection points; and generating the initial obstacle shape based on the plurality of first projection points and the plurality of second projection points.

2. The method according to claim 1, characterized in that The projecting, based on the first projection parameters, a plurality of first vertices used to constitute the first obstacle shape onto the second obstacle shape to obtain a plurality of first projection points includes: generating a plurality of second edges for constituting the second obstacle shape based on a plurality of second vertices for constituting the second obstacle shape; Determine a target edge among the plurality of second edges based on a distance between the first vertex and the second edge, wherein the distance between the first vertex and the target edge is smaller than a distance between the first vertex and other edges, and the other edge is any edge among the plurality of second edges except the target edge; Based on the first projection parameters, the first vertex is projected onto the target edge to obtain the first projection point.

3. The method according to claim 2, characterized in that The projecting the first vertex to the target edge based on the first projection parameter to obtain the first projection point includes: Determine an intersection point between the first vertex and the target edge, wherein a straight line formed by the first vertex and the intersection point is perpendicular to the target edge; generating a direction vector based on the coordinates of the first vertex and the coordinates of the intersection point; The coordinates of the first projection point are obtained based on the direction vector, the first projection parameter and the coordinates of the first vertex.

4. The method according to claim 1, wherein The smoothing process on the initial obstacle shape to obtain the target obstacle shape includes: determining the number of initial vertices used to form the initial obstacle shape; In response to the number of the initial vertices being less than or equal to a first preset number, determining the initial obstacle shape as the target obstacle shape; In response to the number of the initial vertices being greater than a first preset number, the initial vertices are filtered to obtain filtered vertices, and the target obstacle shape is generated based on the filtered vertices.

5. The method according to claim 4, characterized in that The filtering of the initial vertices to obtain filtered vertices includes: Determining a plurality of boundary points of the initial vertex based on the coordinates of the initial vertex, wherein different boundary points are used to represent boundaries of the initial obstacle shape in different directions; Determining an initial vertex between two adjacent boundary points to obtain at least one target pinch point, wherein the two adjacent boundary points are used to represent boundaries of the initial obstacle shape in two adjacent directions; Filtering the at least one target pinch point to obtain filtered pinch points, wherein the number of the filtered pinch points is less than or equal to a second preset number; The filtering vertex is obtained based on the filtering pinch point and the plurality of boundary points.

6. The method according to claim 5, characterized in that The filtering of the at least one target pinch point to obtain a filtered pinch point includes: In response to the at least one target pinch point including a target pinch point, determining the target pinch point as the filter pinch point; In response to the at least one target pinch point including a plurality of target pinch points, the plurality of target pinch points are filtered to obtain the filtered pinch points.

7. The method according to claim 6, characterized in that The filtering of the plurality of target pinch points to obtain the filtered pinch points includes: constructing a boundary line based on the coordinates of the two adjacent boundary points; Determine a projection line of any target grip point on the boundary line, wherein the projection line passes through the target grip point and is perpendicular to the boundary line; The multiple target pinch points are filtered based on the length of the projection line to obtain the filtered pinch point, wherein the length of the projection line corresponding to the filtered pinch point is greater than the length of the projection line corresponding to other pinch points, and the other pinch points are any target pinch points among the multiple target pinch points except the filtered pinch point.

8. The method according to any one of claims 1 to 7, characterized in that The fusing the first obstacle shape in the first obstacle data with the second obstacle shape in the second obstacle data to obtain an initial obstacle shape includes: determining a number of first vertices used to form the first obstacle shape; In response to the number of the first vertices being greater than or equal to a third preset number, the first obstacle shape and the second obstacle shape are merged to obtain the initial obstacle shape.

9. The method according to any one of claims 1 to 7, characterized in that The determining of the fusion parameter based on the obstacle type in the first obstacle data includes: In response to the obstacle type being a preset type, determining that the fusion parameters include a first projection parameter and a second projection parameter.

10. An unmanned vehicle, characterized in that: include: a matching module for matching first obstacle data with second obstacle data in response to the work vehicle detecting the first obstacle data, wherein the first obstacle data is obstacle data identified based on the perception data of the work vehicle at a current moment, and the second obstacle data is obstacle data of obstacles identified based on historical perception data of the work vehicle at a historical moment; a determination module, configured to determine, in response to a successful match between the first obstacle data and the second obstacle data, a fusion parameter based on an obstacle type in the first obstacle data; a fusion module, configured to fuse and smooth the first obstacle shape in the first obstacle data and the second obstacle shape in the second obstacle data based on the fusion parameters to obtain an initial obstacle shape; a smoothing module, configured to smooth the initial obstacle shape to obtain a target obstacle shape; an updating module, configured to update the second obstacle data based on the target obstacle shape; The fusion parameters include a first projection parameter and a second projection parameter, the first projection parameter being used to represent a weight for fusing the first obstacle shape, and the second projection parameter being used to represent a weight for fusing the second obstacle shape. The fusion module is further configured to project, based on the first projection parameter, a plurality of first vertices constituting the first obstacle shape onto the second obstacle shape to obtain a plurality of first projection points; and, based on the second projection parameter, reversely project a plurality of second vertices constituting the second obstacle shape onto the first obstacle shape to obtain a plurality of second projection points. The initial obstacle shape is generated based on the plurality of first projection points and the plurality of second projection points.

Images