Construction machine anti-collision method, device, construction machine, medium, and program product

Abstract

The present application relates to the technical field of engineering machinery, and in particular to an engineering machinery anti-collision method and device, an engineering machinery, a medium and a program product. In a preset work scenario, initial point cloud data collected by a laser radar installed on each support leg of a target engineering machinery is obtained; quality detection is performed on the initial point cloud data corresponding to each support leg to determine whether the initial point cloud data corresponding to each support leg meets a preset quality condition; according to the detection result, a ground pitch angle corresponding to the target engineering machinery is calculated; based on the ground pitch angle, an equivalent ground plane is calculated; and based on the equivalent ground plane, the engineering machinery is subjected to anti-collision processing. The clustering filtering, hierarchical anti-collision control, output of a warning, deceleration and emergency stop instruction are combined; both the anti-collision detection accuracy in a conventional working condition and the stable implementation of anti-collision protection in an extreme working condition with rainwater accumulation are ensured, and the safety of automatic gantry crane operation in rainy days is improved.

Description

Technical Field

[0001] This invention relates to the field of engineering machinery technology, specifically to engineering machinery anti-collision methods, devices, engineering machinery, media, and program products. Background Technology

[0002] In port container yards, automated RTG (Rotating Terrain-Tired Gantry Crane) collision avoidance generally relies on lidar to collect three-dimensional point clouds. By fitting the ground point cloud of the local radar, a ground plane reference is obtained, which is used to separate ground points from obstacle point clouds, thereby realizing obstacle recognition and large vehicle collision avoidance control.

[0003] In some abnormal operating scenarios (such as rainy or snowy weather), a large amount of ground point cloud data collected by lidar equipment is lost. For example, in rainy weather, water accumulation on the storage area creates a specular reflection effect. When the laser emitted by the lidar shines on the water surface, most of the laser energy deviates from the radar's receiving field of view along the specular reflection direction, resulting in the loss of effective ground echoes and a large amount of ground point cloud data.

[0004] In existing technologies, ground plane fitting is completed entirely by relying on the local ground point cloud of a single radar. The loss of a large amount of ground point cloud data will directly cause the ground plane fitting to fail. Once the ground reference fails, the system cannot accurately complete the ground-obstacle point cloud segmentation. Low obstacles such as low road barriers and small debris are easy to miss, and water reflection noise is easy to be misidentified as obstacles, resulting in false alarms and false detections, which seriously threaten the safety of automated gantry crane trolley travel operations.

[0005] Therefore, ensuring the safety of automated gantry crane trolley travel operations under abnormal operating conditions has become an urgent problem to be solved. Summary of the Invention

[0006] This invention provides a method, device, machinery, medium, and program product for preventing collisions with engineering machinery, in order to solve the problem of ensuring the safety of automated gantry crane trolley travel operations under abnormal operating scenarios.

[0007] In a first aspect, the present invention provides a collision avoidance method for engineering machinery, applied to a target engineering machinery. The method includes: acquiring initial point cloud data collected by lidar installed on each leg of the target engineering machinery under a preset operating scenario; performing quality inspection on the initial point cloud data corresponding to each leg to determine whether the initial point cloud data corresponding to each leg meets preset quality conditions; calculating the ground pitch angle corresponding to the target engineering machinery based on the inspection results; calculating an equivalent ground plane based on the ground pitch angle; and performing collision avoidance treatment on the engineering machinery based on the equivalent ground plane.

[0008] The collision avoidance method for construction machinery provided in this application acquires initial point cloud data collected by LiDARs installed on each leg of the target construction machinery under a preset operating scenario. By distributing LiDARs across multiple legs, multi-source redundant point cloud data is obtained. This provides a hardware data foundation for calling radar data from opposite legs when local point cloud fails, overcoming the shortcomings of single-radar sensing data sources and lack of backup in water accumulation conditions.

[0009] The initial point cloud data corresponding to each outrigger is quality inspected to determine whether it meets the preset quality conditions. The outriggers quantitatively evaluate the radar point cloud quality, accurately identifying which radars are disabled due to water reflection interference and which remain usable. This provides a basis for subsequent graded selection of elevation angle data sources, avoiding the direct use of failed point clouds in calculations and reducing the risk of false positives and false negatives from the source. Based on the inspection results, the ground elevation angle corresponding to the target construction machinery is calculated. This achieves radar redundancy priority and structural backup, balancing angle accuracy and availability under extreme conditions, avoiding the accuracy waste caused by directly using structural fitting without considering the working conditions. Based on the ground elevation angle, an equivalent ground plane is calculated. Without relying on local radar, an effective surface point cloud is obtained. The elevation angle is used to perform attitude correction on the initial calibration plane, virtually reconstructing the equivalent ground plane. This addresses the pain points of local ground point cloud loss and ground reference fitting failure caused by water accumulation, providing a reliable ground-obstacle segmentation reference. Based on the equivalent ground plane, collision avoidance measures are implemented for the construction machinery. Obstacle segmentation and recognition are achieved by relying on the equivalent ground plane. Combined with clustering filtering and hierarchical collision avoidance control, warning, deceleration and emergency stop commands are output. It not only ensures the accuracy of collision avoidance detection under normal working conditions, but also can stably achieve collision avoidance protection under extreme working conditions such as rain and water accumulation, thereby improving the safety of automated gantry crane operation in rainy weather.

[0010] In one optional implementation, the initial point cloud data corresponding to each outrigger is subjected to quality inspection to determine whether the initial point cloud data corresponding to each outrigger meets preset quality conditions. This includes: for the initial point cloud data corresponding to each outrigger, determining the ground region of interest corresponding to the LiDAR based on the installation height, installation elevation angle, and geometric parameters of the engineering machinery; based on the ground region of interest, selecting a subset of ground candidate points falling within the ground region of interest from the initial point cloud data; calculating the number of effective ground points, ground point cloud continuity, and the mean and variance of the echo intensity distribution corresponding to the subset of ground candidate points; if the number of effective ground points is greater than a preset effective point number threshold, and the ground point cloud continuity is greater than a preset continuity threshold, and the mean distribution is greater than a preset mean distribution threshold, and the variance distribution is less than a preset variance distribution threshold, then the initial point cloud data corresponding to the outrigger is determined to meet the preset quality conditions.

[0011] The collision avoidance method for engineering machinery provided in this application determines the ground region of interest (ROI) corresponding to the lidar based on the initial point cloud data for each outrigger, the installation height and elevation angle of the lidar corresponding to the outrigger, and the geometric parameters of the engineering machinery. The ground acquisition space range of the lidar is delineated by combining the actual installation parameters of the equipment and the geometric features of the machine body, precisely limiting the search area for ground points; interference from non-ground space-irrelevant point clouds is eliminated, narrowing the data range for subsequent calculations, reducing computational overhead, and improving the accuracy of ground candidate point extraction. Based on the ground ROI, a subset of ground candidate points falling within the ground ROI is selected from the initial point cloud data. Potential ground points are extracted from the original point cloud, and non-ground points such as structures and obstacles are removed, with quality assessment only performed on ground-related points; this avoids interference from non-ground target points in the point cloud quality evaluation, ensuring that quality detection focuses only on the true state of the ground echo. The number of effective ground points corresponding to the ground candidate point subset, the continuity of the ground point cloud, and the mean and variance of the echo intensity distribution are calculated. Quantitative evaluation indicators are extracted from multiple dimensions, including the number of points, spatial geometry, and echo reflection characteristics, moving away from relying on a single indicator to judge the quality of the point cloud. These multi-dimensional features can comprehensively characterize the point cloud degradation caused by water accumulation and specular reflection, providing sufficient quantitative basis for subsequent quality judgment. If the number of valid ground points exceeds a preset threshold, the continuity of the ground point cloud exceeds a preset threshold, the mean distribution exceeds a preset threshold, and the variance is less than a preset threshold, then the initial point cloud data corresponding to the outrigger meets the preset quality conditions. A multi-condition AND joint judgment logic is adopted; only when multiple indicators simultaneously meet the criteria is the point cloud quality deemed acceptable. This avoids misjudgments caused by fluctuations in a single indicator, reliably distinguishing between normal ground echoes and point cloud failures caused by water accumulation and specular reflection, accurately identifying the usability of the ground point cloud for each outrigger radar, and providing reliable judgment results for subsequent graded selection of elevation angle data sources.

[0012] In one optional implementation, the ground pitch angle corresponding to the target construction machinery is calculated based on the detection results, including: if the initial point cloud data corresponding to one outrigger of the target construction machinery does not meet the preset quality conditions, the ground pitch angle corresponding to the target construction machinery is calculated based on the initial point cloud data corresponding to the other outrigger of the target construction machinery.

[0013] The collision avoidance method for engineering machinery provided in this application, if the initial point cloud data corresponding to one outrigger of the target engineering machinery does not meet the preset quality conditions, calculates the ground elevation angle corresponding to the target engineering machinery based on the initial point cloud data corresponding to the other outrigger of the target engineering machinery. It fully utilizes the hardware redundancy resources of the RTG multi-outrigger radar. When the ground point cloud fails due to the specular reflection of water on one outrigger, the point cloud data of the intact outrigger on the opposite side is directly reused to calculate the ground elevation angle, without directly switching to the indirect fitting mode of the structure. It prioritizes the use of measured data directly observed by the radar, preserving the angle calculation accuracy to the greatest extent and avoiding the loss of detection accuracy caused by premature use of structure fitting. It achieves smooth fault tolerance and degradation under mild water accumulation conditions. When only one radar fails, there is no need to switch to the fallback structure calculation mode; reliable ground elevation angles can be obtained using other radars on the same machine. This solves the problem of single-sided radar failure due to water accumulation while ensuring the detection performance of the collision avoidance system. It provides an intermediate layer for the entire graded fault tolerance mechanism. A progressive logic of "local radar available → outboard radar → structure as backup" is formed, which avoids degrading to a lower-precision backup solution when a local radar is interfered with, thus balancing operational accuracy and anti-interference capability. The original elevation angle calculation algorithm remains unchanged. The elevation angle output by the outboard radar has the same physical meaning as the elevation angle output by the local radar. Subsequent equivalent ground plane, obstacle segmentation, and collision avoidance processing can reuse the original algorithm flow, reducing algorithm modifications and improving system engineering feasibility.

[0014] In one optional implementation, collision avoidance processing is performed on the construction machinery based on the detection results, including: if the initial point cloud data corresponding to each outrigger of the target construction machinery does not meet the preset quality conditions, then a rigid structure point cloud set is extracted from the initial point cloud data corresponding to each outrigger; based on the rigid structure point cloud set, the pitch angle of the rigid structure itself corresponding to the rigid structure point cloud set is calculated; and the pitch angle of the rigid structure itself is converted to obtain the ground pitch angle corresponding to the target construction machinery.

[0015] The collision avoidance method for engineering machinery provided in this application extracts a rigid structure point cloud set from the initial point cloud data corresponding to each outrigger of the target engineering machinery if the initial point cloud data for each outrigger does not meet the preset quality conditions. In extreme conditions where all outrigger radars are interfered with by water surface reflection, all ground point clouds are invalid, and no radar-measured ground data source is available, the inherent rigid structures in the excavation yard serve as attitude reference sources. These structures are unaffected by water surface reflection and can provide stable and reliable point cloud input, solving the problem of no attitude reference data source under extreme water conditions and providing a fallback algorithm. Based on the rigid structure point cloud set, the pitch angle of the rigid structure itself corresponding to the point cloud set is calculated. The pitch angle of the structure is calculated based on the fixed geometric features of the rigid structure, bypassing the invalid ground echo point cloud. Multi-source fusion can also be performed using multiple structure point clouds, reducing attitude calculation errors caused by single-point noise and obtaining a more reliable observed attitude of the structure. The pitch angle of the rigid structure itself is converted to obtain the ground pitch angle corresponding to the target engineering machinery. By utilizing the inherent parallel / perpendicular geometric constraints between the structure and the yard ground, the pitch angle of the structure is converted into the ground pitch angle, indirectly obtaining the pitch attitude of the vehicle body relative to the ground. The output ground pitch angle has the same physical meaning as the pitch angle directly observed by radar. The subsequent equivalent ground plane construction, obstacle segmentation, and collision avoidance process can completely reuse the original algorithm logic without redesigning the entire calculation link, ensuring that the collision avoidance service can continue to operate under extreme water accumulation conditions.

[0016] In one optional implementation, the rigid structure point cloud set is extracted from the initial point cloud data corresponding to each outrigger, including: filtering the initial point cloud data corresponding to each outrigger to generate a filtered point cloud dataset; delineating the spatial region of interest of the rigid structure in the global coordinate system based on the equipment size and prior information of the yard layout corresponding to the target engineering machinery; extracting candidate rigid structure point clouds that fall into the spatial region of interest from the filtered point cloud dataset; performing geometric shape screening and echo intensity screening on the candidate rigid structure point clouds, and constraining them based on the preset size library of yard structures to obtain a backup rigid structure point cloud set; performing clustering processing on the backup rigid structure point clouds to generate at least one independent structure point cluster; and verifying each independent structure point cluster to obtain the rigid structure point cloud set.

[0017] The collision avoidance method for engineering machinery provided in this application involves filtering the initial point cloud data corresponding to the outriggers to generate a filtered point cloud dataset. This filtering process removes stray noise from raindrops, water reflections, and spatial outliers, suppressing interference noise in rainy scenes. Simultaneously, it retains valid point clouds of the structural entities, improving the reliability of subsequent point cloud segmentation and reducing noise errors in subsequent attitude calculations. Based on the prior information of the equipment size and yard layout of the target engineering machinery, a spatial region of interest (ROI) for the rigid structure is defined in the global coordinate system. Prior knowledge of the yard is used to limit the target search space, filtering out invalid point clouds in ground areas and distant scenes, narrowing the range of data to be processed; reducing interference from irrelevant point clouds, decreasing the computational load of the algorithm, and improving the computational efficiency of structure extraction. Candidate rigid structure point clouds falling within the ROI are extracted from the filtered point cloud dataset. Potential structure points within the spatial range are quickly selected, completing the initial coarse extraction; irrelevant points not within the ROI are directly excluded, reducing the data processing pressure in subsequent filtering and clustering stages. The candidate rigid structure point cloud is filtered by geometric shape and echo intensity, and constrained by a pre-defined size library of yard structures to obtain a backup rigid structure point cloud. Multi-dimensional filtering based on geometric shape, echo reflection characteristics, and actual object size eliminates scattered virtual points and non-structure interference targets; this narrows the candidate set, reduces the probability of false point clusters generated in subsequent clustering, and improves the confidence of the backup point cloud. The backup rigid structure point cloud is then clustered to generate at least one independent structure point cluster. Clustering divides the point cloud into independent structure objects according to spatial relationships, achieving separation of different structures and enabling attitude calculations for individual structures, providing a foundation for subsequent multi-structure angle fusion. Each independent structure point cluster is validated to obtain the rigid structure point cloud. The point clusters are checked for point quantity, size, and attitude constraints to eliminate false targets formed by noise; deformed and incomplete invalid point clusters are filtered out to output a highly reliable rigid structure point cloud, avoiding the input of inferior point clouds into the attitude calculation process, preventing the obtaining of incorrect pitch angles, and ensuring the accuracy of ground pitch angle calculation under extreme water accumulation conditions.

[0018] In one optional implementation, the initial point cloud data corresponding to each outrigger is filtered to generate a filtered point cloud dataset, including: uniformly converting the initial point cloud data corresponding to each outrigger to the global coordinate system of the target engineering machinery; merging the coordinate-aligned initial point cloud data to obtain merged initial point cloud data; downsampling the merged initial point cloud data to obtain downsampled point cloud data; for each downsampled data point in the downsampled point cloud data, calculating the number of neighboring points in a preset neighborhood area corresponding to the downsampled data point; if the number of neighboring points is less than a preset neighboring point number threshold, deleting the downsampled data point; if the number of neighboring points is greater than or equal to the preset neighboring point number threshold, detecting whether the downsampled data point falls into the ground region of interest based on the data coordinates corresponding to the downsampled data point; if... If a sampled data point falls within a region of interest on the ground, the sampled data point is deleted. If the sampled data point does not fall within a region of interest, the height data corresponding to the sampled data point is obtained. If the height data is less than the preset minimum height of a rigid structure, the sampled data point is marked as a suspected floating point. If the height data is greater than or equal to the preset minimum height of a rigid structure, the sampled data point is marked as a high-confidence protection point of the structure. For each suspected floating point, the single-point echo intensity corresponding to the suspected floating point and the variance of the echo intensity corresponding to all sampled data points in the preset neighborhood of the suspected floating point are calculated. If the single-point echo intensity is less than a preset echo intensity threshold, and / or the echo intensity variance is greater than a preset echo intensity variance threshold, the suspected floating point is deleted. The remaining data points are determined as the filtered point cloud dataset.

[0019] The engineering machinery collision avoidance method provided in this application converts the initial point cloud data corresponding to each outrigger to the global coordinate system of the target engineering machinery. The initial point cloud data after coordinate alignment are then merged to obtain merged initial point cloud data. This eliminates the differences in local coordinate systems of different outrigger lidars, unifying the point clouds of multiple radars into the RTG global coordinate system, achieving spatial alignment of multi-source point clouds, avoiding calculation errors caused by coordinate inconsistencies in subsequent filtering and segmentation, and providing a unified data foundation for multi-radar joint processing. The merged initial point cloud data is downsampled to obtain downsampled point cloud data. This reduces the total number of point clouds while preserving the geometric contour features of the object, reducing the computational overhead of subsequent steps, improving the real-time performance of the algorithm, and meeting the real-time response requirements of the large vehicle collision avoidance system. For each downsampled data point in the downsampled point cloud data, the number of neighboring points within a preset neighborhood area corresponding to the downsampled data point is calculated. The number of neighboring points quantifies the spatial dispersion of the points, providing a quantitative basis for identifying isolated noise points. If the number of neighboring points is less than a preset threshold, the downsampled data point is deleted. The process removes isolated outliers and eliminates scattered invalid points generated by distant floating particles and random scattering, reducing noise interference in subsequent stages. If the number of neighboring points is greater than or equal to a preset threshold, the sampled data point is checked to see if it falls within the ground region of interest (ROI) based on its coordinates. Ground and non-ground points are distinguished, and valid entity points are categorized and branched to prevent ground points from participating in the floating suspicious point identification logic. If a sampled data point falls within the ground ROI, it is deleted. Ground points are directly removed to reduce the amount of data extracted from subsequent structures and to avoid interference from ground points in rigid structure segmentation. If a sampled data point does not fall within the ground ROI, its height data is obtained. Height is introduced as a discrimination criterion to divide the floating space from the space where the structure is located, providing a basis for distinguishing between floating noise and structure entity points. If the height data is less than the preset minimum height for rigid structures, the sampled data point is marked as a floating suspicious point. Points below the minimum height of the structure and in non-ground areas are marked as objects to be checked. Most raindrop and water droplet specular reflections are concentrated in this area, enabling targeted screening of interference points. If the height data is greater than or equal to the preset minimum height of the rigid structure, the sampled data points are marked as high-confidence protection points for the structure. Points within the structure's height range are protected to prevent them from participating in the noise removal logic, thus avoiding the misfiltering of valid points from structures such as containers, guardrails, and columns, ensuring the complete preservation of structure features. For each suspected suspended point, the echo intensity of the corresponding single point and the variance of the echo intensity of all sampled data points within the preset neighborhood of the suspected suspended point are calculated. Combining the single-point echo intensity and the local echo intensity variance, the signal characteristics of raindrop noise and water droplet specular reflection noise are characterized, providing multi-dimensional statistical basis for noise identification.If the echo intensity of a single point is less than a preset echo intensity threshold, and / or the echo intensity variance is greater than a preset echo intensity variance threshold, then the floating suspicious point is deleted. Floating virtual points caused by raindrop scattering and water surface reflection are filtered out using combined echo intensity features, reducing false alarm sources for subsequent obstacle recognition and attitude calculation. The remaining data points are determined as the filtered point cloud dataset. The output is a high-quality point cloud that has undergone multiple rounds of layered filtering, removing various interference noises from the rainy environment while fully preserving the effective point clouds of rigid structures and real obstacles, providing reliable input for subsequent structure point cloud extraction and attitude calculation.

[0020] In one optional implementation, collision avoidance processing of construction machinery is performed based on an equivalent ground plane, including: calculating the vertical distance between each initial point cloud in each initial point cloud data and the equivalent ground plane; if the vertical distance is greater than a preset vertical distance threshold, the initial point cloud is determined as an obstacle candidate point; based on each obstacle candidate point, an obstacle candidate point cloud set is generated; the obstacle candidate point cloud set is clustered to generate at least one candidate obstacle cluster block; each candidate obstacle cluster block is verified to obtain at least one target real obstacle; and the key feature parameters required for collision avoidance decision-making relative to each target real obstacle are calculated; the key feature parameters required for collision avoidance decision-making include the relative distance and relative position between the target construction machinery and the target real obstacle, as well as the target size corresponding to the target real obstacle. Based on the key feature parameters required for collision avoidance decision-making, a collision avoidance strategy corresponding to the actual obstacle is generated; based on the collision avoidance strategy, the engineering machinery is controlled to perform collision avoidance processing.

[0021] The engineering machinery collision avoidance method provided in this application calculates the vertical distance between each initial point cloud and the equivalent ground plane. Using the equivalent ground plane obtained through pitch angle correction as the segmentation benchmark, the height of each point relative to the virtual ground is quantified by the vertical distance, eliminating reliance on invalid local ground point clouds and providing a unified mathematical basis for the division of ground points and obstacle points. If the vertical distance is greater than a preset vertical distance threshold, the initial point cloud is identified as an obstacle candidate point. Points higher than the ground benchmark are quickly separated based on the height threshold, potential obstacle points are screened, and the initial segmentation of ground points and obstacle candidate points is completed, reducing the data scale for subsequent processing. Based on each obstacle candidate point, an obstacle candidate point cloud set is generated. All potential obstacle points are collected to form a unified processing object, facilitating subsequent clustering operations; all point cloud information of suspected targets is centrally retained to avoid the loss of effective obstacle points. The obstacle candidate point cloud set is clustered to generate at least one candidate obstacle cluster block. Spatially adjacent discrete candidate points are aggregated into independent candidate obstacle clusters, transforming the scattered point cloud into object-level candidate objects, thus realizing the point-to-object conversion and providing a foundation for object dimension verification. Each candidate obstacle cluster is verified to obtain at least one real target obstacle. The validity of point clusters is verified, filtering out false clusters formed by residual noise from water accumulation and reflective virtual points, eliminating false alarms; and identifying real obstacles to reduce the probability of false alarms in the collision avoidance system. The key feature parameters required for collision avoidance decisions relative to each real target obstacle are calculated; these parameters include the relative distance and position between the target machinery and the real target obstacle, as well as the target size corresponding to the real obstacle. The necessary quantitative parameters for collision avoidance decisions are extracted, obtaining not only distance information but also orientation and object size, enabling the determination of whether the obstacle is within the vehicle's driving lane, providing a complete and objective calculation basis for the hierarchical collision avoidance strategy. Based on the key feature parameters required for collision avoidance decisions, a collision avoidance strategy corresponding to the real target obstacle is generated. Based on the actual hazard level of the obstacle, a corresponding collision avoidance strategy is matched, and the engineering machinery is controlled to perform collision avoidance actions based on the collision avoidance strategy. Different levels of control strategies, such as warning, deceleration, and emergency stop, are output to achieve graded collision avoidance protection for RTG large vehicles under extreme working conditions such as rain and water accumulation, ensuring the safety of automated operations.

[0022] Secondly, the present invention provides an anti-collision device for engineering machinery, applied to target engineering machinery, the device comprising: The acquisition module is used to acquire the initial point cloud data collected by the lidar installed on each leg of the target construction machinery under a preset operation scenario; The determination module is used to perform quality inspection on the initial point cloud data corresponding to each leg and determine whether the initial point cloud data corresponding to each leg meets the preset quality conditions. The first calculation module is used to calculate the ground pitch angle corresponding to the target engineering machinery based on the detection results; The second calculation module is used to calculate the equivalent ground plane based on the ground pitch angle; The processing module is used to perform collision avoidance processing on engineering machinery based on the equivalent ground plane.

[0023] Thirdly, the present invention provides an engineering machinery, including: an engineering machinery body and an electronic device; the electronic device includes a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the engineering machinery anti-collision method of the first aspect or any corresponding embodiment described above.

[0024] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the engineering machinery collision avoidance method described in the first aspect or any corresponding embodiment thereof.

[0025] Fifthly, the present invention provides a computer program product, including computer instructions, which are used to cause a computer to execute the engineering machinery collision avoidance method described in the first aspect or any corresponding embodiment. Attached Figure Description

[0026] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0027] Figure 1 This is a schematic flowchart of a first method for preventing collisions with engineering machinery according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a second process for a collision avoidance method for engineering machinery according to an embodiment of the present invention; Figure 3 This is a structural block diagram of an engineering machinery anti-collision device according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0030] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0031] According to an embodiment of the present invention, an embodiment of a method for preventing collisions of engineering machinery is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0032] This embodiment provides a collision avoidance method for engineering machinery, which can be used in electronic devices corresponding to the target engineering machinery. Figure 1 This is a flowchart of an engineering machinery collision avoidance method according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps: Step S101: Under the preset operation scenario, acquire the initial point cloud data collected by the lidar installed on each leg of the target construction machinery.

[0033] Among them, the preset operation scenarios mainly refer to the port container yard, including normal operation on sunny days and water accumulation on the yard on rainy days.

[0034] Specifically, lidar is installed on different outriggers of the target engineering machinery. Each lidar is time-synchronized and triggers data acquisition, outputting initial point cloud data. Each initial point cloud data carries the corresponding outrigger number and includes three-dimensional coordinates (x, y, z), echo intensity, and acquisition timestamp information. Different outrigger lidars have independent local coordinate systems, and the acquired data is the point cloud in each lidar's own coordinate system. The distributed arrangement of multiple lidars creates sensing redundancy. If one lidar fails due to water accumulation, the other outrigger lidars may still output valid data. Thus, the electronic equipment can acquire multiple sets of initial point cloud data marked with outriggers and send them to the next step for point cloud quality inspection.

[0035] Step S102: Perform quality inspection on the initial point cloud data corresponding to each leg to determine whether the initial point cloud data corresponding to each leg meets the preset quality conditions.

[0036] Specifically, the electronic device can statistically analyze the initial point cloud data corresponding to each leg, and then compare the statistical results with preset quality conditions. Based on the comparison results, it is determined whether the initial point cloud data corresponding to each leg meets the preset quality conditions.

[0037] This step will be explained in detail below.

[0038] Step S103: Calculate the ground pitch angle corresponding to the target construction machinery based on the detection results.

[0039] Specifically, the electronic equipment can calculate the ground pitch angle of the target construction machinery by using a matching method based on the detection results of the initial point cloud data corresponding to each outrigger.

[0040] This step will be explained in detail below.

[0041] Step S104: Calculate the equivalent ground plane based on the ground pitch angle.

[0042] Specifically, electronic devices can use ground pitch angles for attitude compensation, virtually reconstruct an equivalent ground plane, and provide a ground-obstacle segmentation reference in the event of loss of local ground point clouds.

[0043] Specifically, the electronic device can read the initial ground plane normal vector n0=a0,b0,c0) obtained from the device's factory calibration in the vehicle's horizontal state. Then, the electronic device can use the Y-axis rotation matrix. An attitude rotation correction is applied to the initial ground plane normal vector to obtain an accurate equivalent ground plane normal vector: .in, θ is the calculated ground pitch angle.

[0044] Then, the electronic device can construct the equivalent ground plane equation based on the equivalent normal vector and the calibration intercept D: Ax + By + Cz + D = 0. The electronic device can then load the formula for calculating the vertical distance from a point to the plane: This is used to subsequently calculate the vertical distance of each point cloud relative to the equivalent virtual ground, and finally calculate the complete equivalent ground plane parameters A, B, C, and D, which are then passed to the S105 collision avoidance processing.

[0045] Step S105: Based on the equivalent ground plane, perform anti-collision treatment on the engineering machinery.

[0046] Specifically, electronic devices can identify target obstacles based on an equivalent ground plane, and then perform collision avoidance measures on engineering machinery based on the target obstacles.

[0047] This step will be explained in detail below.

[0048] The collision avoidance method for construction machinery provided in this application acquires initial point cloud data collected by LiDARs installed on each leg of the target construction machinery under a preset operating scenario. By distributing LiDARs across multiple legs, multi-source redundant point cloud data is obtained. This provides a hardware data foundation for calling radar data from opposite legs when local point cloud fails, overcoming the shortcomings of single-radar sensing data sources and lack of backup in water accumulation conditions.

[0049] The initial point cloud data corresponding to each outrigger is quality inspected to determine whether it meets the preset quality conditions. The outriggers quantitatively evaluate the radar point cloud quality, accurately identifying which radars are disabled due to water reflection interference and which remain usable. This provides a basis for subsequent graded selection of elevation angle data sources, avoiding the direct use of failed point clouds in calculations and reducing the risk of false positives and false negatives from the source. Based on the inspection results, the ground elevation angle corresponding to the target construction machinery is calculated. This achieves radar redundancy priority and structural backup, balancing angle accuracy and availability under extreme conditions, avoiding the accuracy waste caused by directly using structural fitting without considering the working conditions. Based on the ground elevation angle, an equivalent ground plane is calculated. Without relying on local radar, an effective surface point cloud is obtained. The elevation angle is used to perform attitude correction on the initial calibration plane, virtually reconstructing the equivalent ground plane. This addresses the pain points of local ground point cloud loss and ground reference fitting failure caused by water accumulation, providing a reliable ground-obstacle segmentation reference. Based on the equivalent ground plane, collision avoidance measures are implemented for the construction machinery. Obstacle segmentation and recognition are achieved by relying on the equivalent ground plane. Combined with clustering filtering and hierarchical collision avoidance control, warning, deceleration and emergency stop commands are output. It not only ensures the accuracy of collision avoidance detection under normal working conditions, but also can stably achieve collision avoidance protection under extreme working conditions such as rain and water accumulation, thereby improving the safety of automated gantry crane operation in rainy weather.

[0050] This embodiment provides a collision avoidance method for engineering machinery, which can be used in electronic devices corresponding to the target engineering machinery. Figure 2 This is a flowchart of an engineering machinery collision avoidance method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Under the preset operation scenario, acquire the initial point cloud data collected by the lidar installed on each leg of the target construction machinery.

[0051] Please refer to the above description of step S101 for details on this step, which will not be repeated here.

[0052] Step S202: Perform quality inspection on the initial point cloud data corresponding to each leg to determine whether the initial point cloud data corresponding to each leg meets the preset quality conditions.

[0053] Specifically, step S202 above may include the following steps: Step S2021: Based on the initial point cloud data corresponding to each outrigger, and according to the installation height, installation pitch angle, and geometric parameters of the LiDAR corresponding to the outrigger, determine the ground region of interest corresponding to the LiDAR.

[0054] Specifically, the electronic equipment can establish a unified vehicle-body reference coordinate system based on the target construction machinery body. The origin of the coordinate system is selected at a preset reference point on the vehicle body; the Y-axis is along the travel direction of the target construction machinery, the Z-axis is vertically upward, and the X-axis is the lateral direction of the target construction machinery. The X-axis represents the installation position coordinate of the outrigger lidar within the vehicle-body coordinate system. lidar ,Y lidar Z lidar The installation height (H) is calculated from the installation height and overall geometric parameters of the radar. inst Corresponding radar Z-axis coordinate Z lidar =H inst All geometric calculations are performed in the vehicle's reference coordinate system to ensure that the output ROI spatial coordinates are consistent with the point cloud coordinate system.

[0055] Then, the electronic equipment can acquire the downward elevation angle α of the radar probe, representing the angle between the radar's main line of sight and the downward direction of the horizontal plane. The radar emits a laser beam that projects downwards, forming an effective illumination area on the actual ground of the storage yard. Combining the radar's maximum detection range Dmax, trigonometric geometric calculations are used to solve for the farthest distance the laser beam can project onto the ground under ideal horizontal ground conditions. The near and far boundaries of the beam are calculated separately, i.e., the closest and farthest projection positions of the radar from the ground, to obtain the coordinate interval [Y] of the region of interest on the ground in the Y direction. min ,Y max ].

[0056] Next, the electronic equipment can read the horizontal field of view parameters of the lidar and, combined with the lidar's installation location, calculate the lateral coverage area of ​​the laser beam. The electronic equipment can then, by considering the geometric constraints of the target construction machinery's overall width, eliminate lateral ranges exceeding the area of ​​interest for the main machine's operation, thus obtaining the X-axis coordinate interval [X...] of the region of interest on the ground. min ,X max ].

[0057] The actual storage yard surface is not a perfectly flat plane; there are slight undulations in the road surface and minor vibrations from vehicles, making it impossible to set a single Z-value. Therefore, the theoretical storage yard surface Z-coordinate is used. ground0 Centered on the axis, allow for height tolerances both upwards and downwards. The coordinate range of the Z-axis is obtained. Among them, height tolerance It is pre-calibrated based on the actual flatness of the yard road surface, which can accommodate minor undulations in the road surface and exclude objects above the ground such as the bottom of guardrails and containers from the Z-section.

[0058] Finally, the intervals of the three coordinate axes X, Y, and Z calculated earlier are combined to form the ground region of interest corresponding to the lidar.

[0059] Step S2022: Based on the region of interest on the ground, select a subset of candidate ground points that fall within the region of interest on the ground from the initial point cloud data.

[0060] Specifically, the electronic device can iterate through each 3D point x in the initial point cloud data corresponding to each leg. i ,y i ,z i Then, it is determined whether the three-dimensional coordinates of the point fall within the ground region of interest obtained in step S2021. All points whose coordinates fall within the ground region of interest are grouped together to obtain a subset of ground candidate points; points not in the ground region of interest are directly discarded and do not participate in the subsequent calculation of ground quality indicators.

[0061] Step S2023: Calculate the number of valid ground points corresponding to the ground candidate point subset, the continuity of the ground point cloud, and the mean and variance of the echo intensity distribution.

[0062] Specifically, the electronic device can perform simple outlier removal on a subset of candidate ground points and then count the total number of valid points within the set to obtain the number of valid ground points. Dry ground has strong diffuse reflection, resulting in a large number of valid ground points; however, rainwater on rainy days has specular reflection, causing most of the laser light to be reflected away from the receiver, leading to a sharp decrease in the number of valid ground points.

[0063] Electronic devices can perform neighborhood connectivity analysis on candidate ground points, and count the proportion of points belonging to continuous connected regions among all candidate ground points as a continuity index. Among them, dry ground has continuous ground points and a high continuity index value; water accumulation interference: the remaining ground points are scattered and fragmented, the connected regions are broken, and the continuity index is significantly reduced.

[0064] The electronic equipment can also extract the echo intensity of each point in a subset of candidate ground points and calculate the arithmetic mean of the echo intensity of all points. Dry cement ground exhibits stable diffuse reflection, resulting in a relatively high mean echo intensity; however, water accumulation and specular reflection result in weaker echo energy and a significantly lower mean.

[0065] Electronic equipment can calculate the variance of the echo intensity at candidate ground points, characterizing the degree of fluctuation in the echo signal. Normal ground echo signals are stable with small variance; under conditions of water reflection, the echo intensity at a few return points fluctuates greatly, and the variance increases significantly.

[0066] Finally, the number of valid ground points corresponding to the ground candidate point subset, the continuity of the ground point cloud, and the mean and variance of the echo intensity distribution were calculated.

[0067] Step S2024: If the number of valid ground points is greater than the preset valid point number threshold, the continuity of the ground point cloud is greater than the preset continuity threshold, the distribution mean is greater than the preset distribution mean threshold, and the distribution variance is less than the preset distribution variance threshold, then the initial point cloud data corresponding to the leg is determined to meet the preset quality conditions.

[0068] Specifically, the electronic device can pre-store four sets of thresholds obtained from calibration: a preset threshold for the number of valid points, a preset threshold for continuity, a preset threshold for the mean of the distribution, and a preset threshold for the variance of the distribution.

[0069] Then, the electronic device can determine whether four conditions are met simultaneously (logical AND): ① number of valid ground points > preset valid point number threshold; ② ground point cloud continuity > preset continuity threshold; ③ echo intensity distribution mean > preset distribution mean threshold; ④ echo intensity distribution variance < preset distribution variance threshold.

[0070] If all the above conditions are met simultaneously, the initial point cloud data corresponding to the outrigger is determined to meet the preset quality conditions, indicating that the ground echo of the outrigger is intact and the point cloud can be used for ground pitch angle calculation.

[0071] If any condition is not met, the initial point cloud data of the outrigger is determined to not meet the preset quality conditions, indicating that the outrigger is affected by the specular reflection of accumulated water, the ground point cloud is invalid, and it cannot be directly used for ground fitting. Finally, the outrigger is determined to meet or not meet the preset quality conditions, and the result is output to S203 for ground pitch angle classification calculation.

[0072] Step S203: Calculate the ground pitch angle corresponding to the target construction machinery based on the detection results.

[0073] Specifically, step S203 above may include the following steps: In one optional implementation, if the initial point cloud data corresponding to each outrigger meets the preset quality conditions, the ground pitch angle corresponding to the target engineering machinery is calculated based on the initial point cloud data corresponding to each outrigger.

[0074] Specifically, the electronic equipment can iterate through the judgment markers of all outriggers to confirm that the initial point cloud data of each outrigger fully meets the preset quality conditions and that all radar ground point clouds are complete and valid. If any outrigger fails to meet the preset quality conditions, the system exits this branch and enters a fault-tolerant branch with the opposite outrigger or structure as a fallback.

[0075] All outrigger point cloud quality is qualified. The electronic equipment can select a single outrigger point cloud on the local operation side, or select multiple outriggers to participate in the calculation.

[0076] Optionally, the electronic device can directly select the initial point cloud of the local support leg, which requires less computation and has a fast response speed.

[0077] Optionally, the electronic equipment can also select multiple qualified outriggers to participate in the calculation, using multi-source observation to suppress single-point noise and further improve angle reliability.

[0078] Specifically, for each selected leg, the initial point cloud data is used to retrieve the pre-calculated region of interest (ROI) on the ground. Points falling within the ROI are selected from the initial point cloud data of that leg, resulting in a subset of candidate ground points for that leg. At this point, due to favorable operating conditions, the candidate ground point subset has a sufficient number of points, continuous point cloud, and excellent echo intensity. Then, the electronic equipment can use a least-squares plane fitting algorithm to fit the candidate ground point cloud of that leg to the ground plane in the radar's local coordinate system. Combining this with the radar's installation attitude calibration parameters, the attitude of the ground plane in the local coordinate system is transformed to the global coordinate system of the target construction machinery vehicle; the ground pitch angle observed by that leg is then calculated. If only a single local leg is selected, a single set of pitch angles is directly obtained, skipping multi-source fusion.

[0079] When multiple qualified outriggers are used for joint calculation, weights are assigned based on the number of ground candidate points for each outrigger and the plane fitting residual; the more ground points and the smaller the fitting residual, the higher the weight is assigned. The pitch angles θ1, θ2…θ1 output by the multiple outriggers are then calculated. n Perform weighted average fusion: Among them, w i Let θ be the weighting coefficient corresponding to the i-th outrigger, and then we can obtain the fused ground pitch angle θ.

[0080] Optionally, the electronic equipment can perform a rationality check on the calculated ground pitch angle θ. The actual pitch angle of the target construction machinery is a small angle; it is determined whether θ is less than the maximum allowable pitch threshold. The historical pitch angle output from the previous frame is compared to determine whether the angle jump in this frame is within the allowable range. The ground plane fitting residual is checked; the residual cannot exceed the residual threshold. If the check passes, a valid ground pitch angle θ is output for equivalent ground plane calculation. If the check fails, the angle calculation is deemed unsuccessful, the angle is abandoned, and the process switches to the rigid structure fallback branch.

[0081] Step S2031: If the initial point cloud data corresponding to one outrigger of the target construction machinery does not meet the preset quality conditions, then the ground pitch angle corresponding to the target construction machinery is calculated based on the initial point cloud data corresponding to the other outrigger of the target construction machinery.

[0082] Specifically, the electronic equipment can read the outrigger quality inspection results to determine the outriggers on the working side (local side) of the target construction machinery. If the initial point cloud data corresponding to the working side (local side) outrigger does not meet the preset quality conditions, it indicates that the local outrigger is affected by the specular reflection of rainwater, resulting in the loss of a large number of effective ground points, making it impossible to calculate the ground pitch angle using the local outrigger. Then, the electronic equipment further verifies the quality markers of other (opposite side) outriggers to confirm that at least one outrigger on the opposite side has an initial point cloud that meets the preset quality conditions. The ground point cloud quality of this outrigger is qualified and can be used for ground attitude calculation. The electronic equipment can prioritize selecting the outrigger with the best point cloud quality index (the largest number of effective ground points and the best continuity) from all available outriggers. If multiple qualified outriggers on the opposite side exist, multiple outrigger point clouds can also be selected.

[0083] Then, the electronic equipment can identify the initial point cloud data of the opposite outrigger and extract the ground point cloud data of the region of interest on the ground of that outrigger. The electronic equipment can then use the intact ground point cloud data of the opposite outrigger to perform ground plane fitting, solving for the ground plane in the local coordinate system of the opposite radar. Based on the installation and calibration parameters of the opposite radar, the fitted ground attitude is transformed to the global coordinate system of the target mechanical engineering vehicle, and the ground pitch angle θ of the vehicle body relative to the actual ground of the storage yard is calculated. This pitch angle is calculated by directly observing the actual ground with the lidar, and is a measured angle with high accuracy, superior to pitch angles indirectly calculated using rigid structures.

[0084] Optionally, the electronic equipment can perform a rationality check on the calculated ground pitch angle θ. The actual pitch angle of the target construction machinery is a small angle; it is determined whether θ is less than the maximum allowable pitch threshold. The historical pitch angle output from the previous frame is compared to determine whether the angle jump in this frame is within the allowable range. The ground plane fitting residual is checked; the residual cannot exceed the residual threshold. If the check passes, a valid ground pitch angle θ is output for equivalent ground plane calculation. If the check fails, the angle calculation is deemed unsuccessful, the angle is abandoned, and the process switches to the rigid structure fallback branch.

[0085] Step S2032: If the initial point cloud data corresponding to each leg of the target engineering machinery does not meet the preset quality conditions, then extract the rigid structure point cloud set from the initial point cloud data corresponding to each leg.

[0086] Specifically, the step S2032 above, "extracting the rigid structure point cloud set from the initial point cloud data corresponding to each leg," may include the following steps: Step a1: Filter the initial point cloud data corresponding to each leg to generate a filtered point cloud dataset.

[0087] Specifically, step a1 above may include the following steps: Step a11: Convert the initial point cloud data corresponding to each outrigger to the global coordinate system of the target engineering machinery, and merge the initial point cloud data after coordinate alignment to obtain merged initial point cloud data.

[0088] Specifically, the initial point cloud data output by each outrigger lidar has coordinates originating from that lidar itself, belonging to the lidar's local coordinate system; the coordinate systems of different outrigger lidars are independent of each other. The electronic equipment can read the pre-calibrated extrinsic parameters of each outrigger lidar: translation vector and rotation matrix. Then, using the extrinsic parameter transformation matrix, it transforms the x-axis of each point under that outrigger. lidar ,y lidar ,z lidar The transformation is performed from the radar local coordinate system to the target engineering machinery's global vehicle coordinate system. The transformation formula is: , where R is the rotation matrix and t is the translation vector.

[0089] After coordinate transformation is completed for all leg point clouds, all points have a unified global coordinate reference, achieving spatial alignment. The aligned point clouds are then merged into a single point cloud container to obtain the merged initial point cloud data.

[0090] Step a12: Downsample the initial point cloud data to obtain downsampled point cloud data.

[0091] Specifically, electronic devices can employ voxel grid downsampling, setting the side length of the cubic grid. Only one representative point is retained within each grid. The initial point cloud data is traversed and merged to complete downsampling, resulting in the downsampled point cloud data. Downsampling does not delete large entities such as structures; it only reduces redundant dense points. Small, scattered noise points from rainy weather are partially filtered out. This reduces the data overhead of subsequent neighborhood searches, statistical calculations, and clustering operations, meeting the real-time constraints of the gantry crane anti-collision system.

[0092] Step a13: For each downsampled data point in the downsampled point cloud data, calculate the number of neighboring points in the preset neighborhood area corresponding to the downsampled data point.

[0093] Specifically, for each downsampled data point in the downsampled point cloud data, the electronic device can traverse every downsampled data point in the downsampled point cloud data. A pre-defined three-dimensional neighborhood region is determined centered on the current downsampled data point. All other points falling within the pre-defined neighborhood region are retrieved, and the number of neighboring points is counted.

[0094] Step a14: If the number of neighboring points is less than the preset threshold for the number of neighboring points, then delete the downsampled data points.

[0095] The electronic device can receive a preset threshold for the number of neighboring points input by the user, or it can receive a preset threshold for the number of neighboring points sent by other devices. The electronic device can also set a preset threshold for the number of neighboring points according to the actual situation.

[0096] Specifically, the electronic device can compare the number of neighboring points with a preset neighboring point number threshold. If the number of neighboring points is less than the preset neighboring point number threshold, the downsampled data point is deleted. If the number of neighboring points is greater than or equal to the preset neighboring point number threshold, the downsampled data point is retained.

[0097] Step a15: If the number of neighboring points is greater than or equal to the preset threshold for the number of neighboring points, then based on the data coordinates corresponding to the downsampled data points, detect whether the downsampled data points fall into the region of interest on the ground.

[0098] Specifically, if the number of neighboring points is greater than or equal to a preset threshold, the electronic device can read the pre-stored X / Y / Z coordinate range of the region of interest (ROI) on the ground of the storage yard. Then, based on the three-dimensional coordinates (x, y, z) of the current point corresponding to the downsampled data point, it detects whether the downsampled data point falls within the range of the ROI on the ground.

[0099] Step a16: If the downsampled data point falls into the region of interest on the ground, then delete the downsampled data point.

[0100] Specifically, if a downsampled data point falls into a region of interest on the ground, the electronic device can delete the downsampled data point.

[0101] Step a17: If the downsampled data point does not fall into the region of interest on the ground, then obtain the height data corresponding to the downsampled data point.

[0102] Specifically, if the downsampled data point does not fall within the region of interest on the ground, the electronic device can read the Z-axis coordinate of the current downsampled data point in the global coordinate system, which is the height data of the point relative to the vehicle body reference.

[0103] In step a18, if the height data is less than the preset minimum height of the rigid structure, the sampled data point is marked as a floating suspicious point.

[0104] Specifically, the electronic equipment can compare height data with a preset minimum height threshold for rigid structures. This threshold is pre-calibrated based on the minimum installation height of structures such as yard fences and columns.

[0105] If the height data is less than the preset minimum height of the rigid structure, the point is determined to be in the low-altitude suspended space between the ground and the structure, and is marked as a suspected suspended point.

[0106] Step a19: If the height data is greater than or equal to the preset minimum height of the rigid structure, then mark the downsampled data point as a high-confidence protection point of the structure.

[0107] Specifically, if the height data is greater than or equal to the preset minimum height of a rigid structure, the sampled data point is determined to be highly likely to belong to the surface of a rigid structure such as a guardrail or column, and is marked as a high-confidence protection point. High-confidence protection points skip the subsequent floating point echo verification process and will not be mistakenly deleted.

[0108] Step a110: For each suspended suspicious point, calculate the single-point echo intensity corresponding to the suspended suspicious point and the variance of the echo intensity corresponding to all downsampled data points in the preset neighborhood of the suspended suspicious point.

[0109] Specifically, the electronic device can traverse all data marked as floating suspicious points and obtain the single-point echo intensity value Ip of each floating suspicious point. Then, the electronic device can define a preset neighborhood range centered on the floating suspicious point, extract the echo intensity of all points within the neighborhood, and calculate the variance of the neighborhood echo intensity. Among them, the virtual points of water accumulation and specular reflection, and the points of raindrop scattering, have weak echo energy and low intensity at single points; the signal fluctuations within the neighborhood are large, and the variance is high. The real obstacles at low altitudes have stable echo intensity and small neighborhood variance.

[0110] Step a111: If the single-point echo intensity is less than the preset echo intensity threshold, and / or the echo intensity variance is greater than the preset echo intensity variance threshold, then the floating suspicious point will be deleted.

[0111] Specifically, the electronic device can receive preset echo intensity threshold and preset echo intensity variance threshold input by the user, or it can receive preset echo intensity threshold and preset echo intensity variance threshold sent by other devices, or it can set preset echo intensity threshold and preset echo intensity variance threshold according to the actual situation.

[0112] The electronic device can then compare the single-point echo intensity with a preset echo intensity threshold and the echo intensity variance with a preset echo intensity variance threshold. If the single-point echo intensity is less than the preset echo intensity threshold and / or the echo intensity variance is greater than the preset echo intensity variance threshold, the floating suspicious point will be deleted.

[0113] If the echo intensity of a single point is greater than or equal to the preset echo intensity threshold, and the echo intensity variance is less than or equal to the preset echo intensity variance threshold, then the floating suspicious point is retained.

[0114] Step a112: The retained data points are identified as the filtered point cloud dataset.

[0115] Specifically, the electronic device collects all the remaining points after all the previous filtering, deletion, labeling, and verification steps into a single dataset, which is the filtered point cloud dataset. The dataset consists of: high-confidence protected points of structures and low-altitude real object points retained through echo verification.

[0116] Step a2: Based on the prior information of the equipment size and yard layout corresponding to the target engineering machinery, delineate the spatial region of interest of the rigid structure in the global coordinate system.

[0117] Specifically, the electronic equipment can read the original outline dimensions of the target construction machinery, including the length, width, and height of the vehicle body. A safety margin is preset for expanding the three-dimensional bounding box of the vehicle body outwards, resulting in an expanded bounding box. This expanded bounding box is then designated as a prohibited search area. The target construction machinery's own frame and outriggers will generate a large amount of point cloud data; by using the prohibited search area, the space occupied by the vehicle body is excluded from the structure search range, preventing the vehicle body's own point cloud from being misidentified as a rigid structure outside the storage yard.

[0118] Then, the electronic equipment can retrieve prior information about the yard layout to obtain the lateral position range of structures such as pillars and guardrails on both sides of the track relative to the centerline of the trolley. The prohibited search area for the trolley body is removed on the X-axis; the lateral range of structures existing on the left and right sides of the trolley is retained, resulting in the upper and lower limits of the lateral coordinates. Among them, X struct,min X represents the left lateral boundary of the structure.struct,max This defines the right lateral boundary of the structure. It restricts the structure to appearing only on the left and right sides of the vehicle lane; structures in the middle of the lane are no longer searched, reducing invalid searches.

[0119] Next, the electronic equipment can combine the maximum detection range of the lidar and the warning range for large vehicles to define the effective search range for structures along the direction of vehicle travel (Y-axis). After removing blind spots too close to the vehicle and distant areas beyond the effective observation range of the radar, the upper and lower limits of the Y-axis coordinates are obtained. Among them, Y struct,min Y represents the shortest retrieval distance along the forward direction. struct,max This represents the furthest retrieval distance along the forward direction.

[0120] Finally, the electronic equipment retrieves the prior installation parameters of the rigid structures in the storage yard: the minimum installation height and the maximum height of the structures. This determines the vertical search range. Among them, Z struct,min This is the minimum height for a structure; structures below this height will no longer be searched. struct,max This represents the maximum height of the structure; above this height, there are generally no pillars, railings, or other targets. This Z-interval can filter out residual ground points, low-altitude suspended noise points, and irrelevant high-altitude point clouds.

[0121] The electronic device combines the constraints obtained from the previous X, Y, and Z coordinate axes to form the spatial region of interest of the rigid structure.

[0122] Step a3: Extract the candidate rigid structure point cloud set that falls into the region of interest in space from the filtered point cloud dataset.

[0123] Specifically, the electronic device can traverse every 3D (x, y, y) dimension in the filtered point cloud dataset. i ,y i ,z i The system determines whether the three-dimensional coordinates of a point fall within the defined spatial region of interest (ROI) of the rigid structure. Then, the electronic device can aggregate all points whose coordinates satisfy the ROI conditions of the rigid structure, forming a candidate rigid structure point agglomeration; points falling outside the ROI are directly discarded.

[0124] Step a4 involves screening the candidate rigid structure point set by geometric shape and echo intensity, and constraining it based on the preset size library of yard structures to obtain the standby rigid structure point set.

[0125] Specifically, the electronic device can sequentially extract each three-dimensional point P within the candidate rigid structure point set. i (x i ,y i ,zi Electronic devices can be set to a fixed spherical neighborhood search radius r. nei Search by P i Center of the sphere, radius r nei All neighboring points within the range constitute the local neighborhood set of that point. If the number of points in the neighborhood is too small to meet the minimum neighborhood point requirement, the point is directly identified as noise and removed, and no further morphological calculations are performed.

[0126] Then, for points within the effective neighborhood, the electronic device first solves for the center point of the 3D coordinates of the neighborhood point cloud: , where N represents the total number of points in the neighborhood.

[0127] Electronic devices can construct the covariance matrix of neighborhood point clouds. Then, the electronic device can perform eigenvalue decomposition on the covariance matrix to obtain three eigenvalues ​​sorted from largest to smallest. Electronic devices can determine local morphology based on the proportional relationship between feature values: planar features: The corresponding railings and wall planes should be preserved; linear features: Corresponding to column-like linear structures, these are retained; spherical discrete noise points: The point cloud was scattered and disordered, and was identified as virtual points reflecting water accumulation or raindrop noise, so these points were directly deleted. After filtering, a set of intermediate points with acceptable geometric shapes was obtained.

[0128] Next, the electronic equipment can retrieve the pre-calibrated echo intensity range of the rigid structure. This interval represents the normal reflection signal range of concrete columns and metal railings. After geometric filtering, all remaining points are traversed, and the echo intensity attribute Ip of each point is read. If... If the echo intensity is normal, retain that point; if If the echo energy is weak, it is likely a virtual point reflected by the water-filled mirror surface and should be eliminated; if If the signal exhibits an abnormal jump, it is identified as an abnormal scattering point and discarded. After screening, a set of intermediate points that pass both geometric and echo verification is obtained.

[0129] Next, the electronic equipment can read a pre-built offline database of preset dimensions for yard structures, which stores the theoretical width, thickness, and height ranges for targets such as columns, guardrails, and track stops. A temporary 3D bounding box is generated for the current set of intermediate points, and the dimensions of the bounding box in the X / Y / Z directions are calculated: W x W y H zThe electronic device compares the bounding box size with the size ranges of various structures in the size library. If the bounding box size falls within the allowable range of any type of structure's size: the point set size is reasonable and it is retained. If it contains only a very small number of points and the bounding box size is much smaller than the minimum specification of the structure: it is judged as a scattered noise cluster and is removed entirely. If the bounding box size is much larger than the maximum theoretical size of the structure: it belongs to a large irrelevant scene and is removed. This step is a coarse prediction, only filtering out small clusters of points that are obviously abnormal. The complete verification at the object level is postponed until after clustering.

[0130] Finally, the electronic equipment will collect all the three-dimensional points that have been filtered by geometry, echo intensity, and yard structure size constraints into the same point cloud container to form a new point cloud set, which is the standby rigid structure point cloud set.

[0131] Step a5: Cluster the standby rigid structure point cloud to generate at least one independent structure point cluster.

[0132] Specifically, the electronic device can use a Euclidean clustering algorithm and set a clustering distance threshold. When the spatial distance between two points is less than the clustering threshold, the two points are determined to belong to the same object. The electronic device can traverse the point cloud of the spare rigid structures to complete the clustering segmentation. Points that are spatially far apart and separated from each other will be assigned to different clusters. After clustering, one or more non-overlapping independent structure point clusters are generated; each point cluster represents a potential independent structure target.

[0133] Step a6: Verify each independent structure point cluster to obtain a rigid structure point cloud.

[0134] Specifically, the electronic equipment can determine whether the total number of points within a cluster exceeds a minimum point threshold. If the number of points is too low, it is considered a pseudo-cluster generated by noise and is directly discarded. The electronic equipment can calculate the length, width, and height of the 3D bounding box of the point cluster; it then compares this with a preset size library of yard structures, discarding point clusters whose dimensions deviate significantly from the structure specifications. The electronic equipment can perform planar or cylindrical fitting on the point cluster as a whole and verify the fitting residuals. If the residuals are too large, it indicates that the point clusters are scattered and not flat rigid structures, and they are discarded. Point clusters that pass all verification items are determined to be real and valid structure objects. The qualified point clusters are merged and aggregated to obtain a rigid structure point cloud.

[0135] Step S2033: Based on the point set of rigid structures, calculate the pitch angle of the rigid structure itself corresponding to the point set of rigid structures.

[0136] Specifically, a rigid structure point cluster can contain one or more verified independent structure point clusters. If it is a single structure, the only qualified point cluster is selected directly; if it is multiple structures, the structure with the most points and the smallest plane fitting residual is selected first; alternatively, the attitude of multiple structures can be solved separately, and then multi-source angle fusion is performed.

[0137] For rigid structures such as yard fences, blocks, and column sides, whose surfaces are planes, electronic equipment uses the least squares method to fit the plane equations of the structure surfaces. Let the general plane equation be: The electronic device aims to minimize the sum of squared distances from all points to the plane. It solves for the plane parameters a, b, c, and d to obtain the plane normal vector. .

[0138] If the target is a vertical column-shaped linear structure, then the spatial central axis is fitted instead, the axis direction vector is extracted, and the attitude calculation is carried out instead of the plane normal vector.

[0139] The pitch angle is defined as the angle of inclination of the plane or axis of a structure relative to the horizontal plane around the horizontal X-axis.

[0140] The horizontal plane normal vector (vertically upward) is: The electronic equipment can calculate the angle between the plane normal vector of the structure and the vertical direction, and convert it into the pitch angle of the structure relative to the vehicle coordinate system, denoted as . , which is the pitch angle of a rigid structure itself.

[0141] Optionally, when the pitch angles of multiple structures are calculated simultaneously... The electronic equipment can be weighted using the reciprocal of the plane fitting residuals of each structure; the smaller the residual, the higher the weight. The weighted average yields the pitch angle of the fused structure itself. , ,in, The residual of the plane fitting of the i-th structure; It is a very small constant to prevent the denominator from being zero.

[0142] Step S2034: Convert the pitch angle of the rigid structure itself to obtain the ground pitch angle corresponding to the target engineering machinery.

[0143] Specifically, rigid structures in the storage yard must be constructed and installed strictly perpendicular to the horizontal plane of the storage yard. Under ideal conditions, the pitch angle of the structure itself is zero. .

[0144] When the target construction machinery vehicle pitches, the lidar tilts accordingly; the lidar's view of the originally vertical structure will generate an observed tilt amount. The observed tilt is essentially the apparent angle caused by the tilt of the vehicle body, rather than the tilt of the structure itself.

[0145] After the radar pitches up and down with the target construction machinery, the coordinate system tilts; the radar then observes the apparent pitch angle of the structure. Pitch angle of the vehicle body relative to the ground They are of equal size. The attitude transition equation is: The sign is determined by the Z-axis direction of the coordinate system and the positive direction of the pitch angle, and can be corrected for positive or negative signs based on the calibration results. The structure itself is fixed perpendicular to the ground; the observed tilt angle of the structure is the pitch offset of the vehicle body relative to the horizontal ground.

[0146] Electronic equipment can calculate the ground pitch angle Compared to the previous frame's output, the angle change must not exceed the maximum permissible jump threshold between frames, filtering out instantaneous noise. The target construction machinery's body pitch is a small angle, which is considered... The pitch angle is less than the vehicle's maximum pitch threshold. If the verification passes, the ground pitch angle is considered valid; if the verification fails, the fallback branch angle calculation fails, which may trigger a system alarm or use the historical cached pitch angle.

[0147] Step S204: Calculate the equivalent ground plane based on the ground pitch angle.

[0148] Please refer to the above description of step S104 for details on this step, which will not be repeated here.

[0149] Step S205: Based on the equivalent ground plane, perform anti-collision treatment on the engineering machinery.

[0150] Specifically, step S205 above may include the following steps: Step S2051: Calculate the vertical distance between each initial point cloud in each initial point cloud data and the equivalent ground plane.

[0151] Specifically, the electronic device can traverse every 3D sampling point in the entire initial point cloud data and extract any initial point P. i x i ,y i ,z i ).

[0152] Electronic devices calculate the perpendicular (normal) distance from a point to a plane using a formula. i Vertical distance to the equivalent ground plane: The equivalent ground plane equation is: ax+by+cz+d=0.

[0153] Electronic devices can measure distance value D i With point P i Each element is bound individually to provide data for the next threshold determination.

[0154] Step S2052: If the vertical distance is greater than the preset vertical distance threshold, the initial point cloud is determined as an obstacle candidate point.

[0155] Specifically, the electronic device can receive a preset vertical distance threshold input by the user, or a preset vertical distance threshold sent by other devices. The electronic device then compares the vertical distance with the preset vertical distance threshold. If the vertical distance is greater than the preset vertical distance threshold, the initial point cloud is identified as a candidate obstacle point.

[0156] Step S2053: Generate a set of obstacle candidate point clouds based on each obstacle candidate point.

[0157] Specifically, the electronic device generates a cloud set of obstacle candidate points based on each obstacle candidate point.

[0158] Step S2054: Perform clustering processing on the candidate obstacle point cloud set to generate at least one candidate obstacle cluster block.

[0159] Specifically, electronic devices can set a three-dimensional spatial clustering distance threshold T. cluster If the Euclidean distance between two points is less than a threshold, then the two points are determined to belong to the same object.

[0160] The electronic device can traverse the candidate point cloud set of obstacles and perform clustering segmentation. Points that are spatially close to each other are grouped into the same cluster; points that are spatially far apart are grouped into different clusters. Each cluster represents a potential independent object; ultimately, one or more non-overlapping candidate obstacle clusters are generated. This completes the target instance segmentation, distinguishing multiple obstacles in the scene, which facilitates subsequent obstacle-by-obstacle verification and parameter calculation.

[0161] Step S2055: Verify each candidate obstacle cluster block to obtain at least one target real obstacle.

[0162] Specifically, the electronic device can verify the number of points in each candidate obstacle cluster block. If the total number of points in the cluster is greater than or equal to the minimum number of points threshold, it is judged as a noise pseudo-cluster and is removed.

[0163] The electronic device can perform three-dimensional bounding box verification on each candidate obstacle cluster block, calculate the length, width, and height of the cluster block bounding box, and filter out abnormal blocks that are too small or exceed the radar detection range.

[0164] The electronic device can perform distance consistency checks on each candidate obstacle cluster: the depth deviation between each point within a cluster and the vehicle body cannot be too large, excluding discrete floating virtual point clusters. Clusters that pass all checks are determined to be real target obstacles.

[0165] Step S2056: Calculate the key feature parameters required for collision avoidance decision-making of the target engineering machinery relative to each real obstacle.

[0166] The key characteristic parameters required for collision avoidance decision-making include the relative distance and relative position between the target engineering machinery and the target real obstacle, as well as the target size corresponding to the target real obstacle.

[0167] Specifically, the electronic device can calculate the 3D axis-aligned bounding box of the obstacle cluster blocks corresponding to the target real obstacle: X min ,X max ,Y min ,Y max Z min Z max Solve for the obstacle dimensions: , , Where W is the width, L is the length along the direction of travel of the target engineering machinery, and H is the height of the obstacle. The center point of the obstacle bounding box is: .

[0168] The electronic equipment can use the reference origin of the target engineering machinery body as a reference to solve the three-dimensional Euclidean distance from the origin of the vehicle body to the center point of the obstacle; at the same time, it can extract the longitudinal distance (the most critical distance for collision avoidance) along the forward direction of the vehicle (Y-axis).

[0169] The electronic device can determine whether a real obstacle is located on the left, right, or directly in front of the vehicle based on the center point coordinates, obtaining lateral and longitudinal relative position information. It then obtains key collision avoidance feature parameters such as relative distance, relative position, and target size for each obstacle. This completes the quantitative representation of the obstacle state, transforming point cloud data into structured parameters that the collision avoidance controller can directly use.

[0170] Step S2057: Based on the key feature parameters required for collision avoidance decision-making, generate a collision avoidance strategy corresponding to the target real obstacle.

[0171] Specifically, electronic devices can be divided into multiple safety zones based on their longitudinal relative distance: a safety zone, a warning zone, a deceleration zone, and an emergency stop zone. The danger level gradually increases as the distance decreases.

[0172] For example, an electronic device can pre-calibrate four longitudinal distance thresholds: D1, D2, and D3, satisfying: D1 > D2 > D3 > 0. If d y >D1 is the safe zone, Lv-0 is safe, and the basic strategy is to maintain the original speed; if D2 <d y If D1 is less than or equal to 1, it is a warning zone, Lv-1 is low risk, and the basic strategy is audible and visual warning; if D3 is less than or equal to 1, it is a warning zone. <d y If the value is ≤D2, it is a deceleration zone, and Lv-2 is dangerous; the basic strategy is to reduce speed. If the value is 0... <d y If the value is ≤D3, it is an emergency stop zone. If the value is Lv-3, it is a high-risk zone and the basic strategy is emergency braking.

[0173] Set two independent adjustment weight coefficients: horizontal position weight k x Obstacle size weight k w .

[0174] The overall adjusted weight is: The hazard level correction uses a level-up mechanism: the higher the correction weight, the easier it is to increase the hazard level, triggering more conservative collision avoidance actions in advance and avoiding collision risks.

[0175] Among them, the obstacle lateral position correction factor k x It is used to determine whether an obstacle falls within the driving lane of a large vehicle; obstacles inside the lane pose a higher risk than obstacles outside the lane.

[0176] Let the half-width of the target construction machinery be X. half The horizontal coordinate of the obstacle's center point (x) c Lateral offset of the obstacle from the centerline of the vehicle: .

[0177] For example, regarding the lane center hazard zone (where the obstacle projection overlaps with the large vehicle's travel path): Then k x =1.0, the highest risk, with no reduction in hazard level.

[0178] For the lane near-side edge area: Then k x =0.8, the obstacle is close to the side of the vehicle, the risk is slightly reduced.

[0179] For the outer edge of the lane: k x =0.6, the obstacle is far from the vehicle's passageway, and the risk decreases. Among them, As a lateral safety margin, it is pre-calibrated.

[0180] Among them, the obstacle width dimension correction factor k w The wider the obstacle, the larger the area it crosses the lane, the higher the probability of collision, and the higher the hazard weight. Let the minimum reference width W be... min Maximum saturation width W max .

[0181] For narrow obstacles: Then k w =0.7; For medium-width obstacles: ,but The weight increases linearly with the width of the obstacle; for wide obstacles (completely spanning the lane): Then k w =1.0. The wider the obstacle, the closer kw is to 1, and the less likely the danger level is to decay.

[0182] The electronic device reads the basic hazard level L0 from the longitudinal distance dy. Then, it considers the lateral offset from the obstacle. Obtain the horizontal weight k x The size weight k is obtained based on the obstacle width W. w Finally, calculate the overall weight. .when (Raise the threshold, for example, to 0.9): The danger level is increased by one level (maximum is Lv-3 emergency stop); when The hazard level remains unchanged from the original base level; when (Lower the threshold, for example, take 0.65): The danger level is lowered by one level (lower limit Lv-0 safe zone).

[0183] For example, if the longitudinal distance of the obstacle is in the warning zone Lv-1, and the obstacle is very wide and located in the center of the lane, K=1.0 is greater than 0.9, the level is upgraded by one level to the deceleration zone Lv-2, and the large vehicle decelerates in advance.

[0184] The revised danger level is Lv-0, the safe zone, and the collision avoidance strategy is for the vehicle to maintain its current speed without any intervention; the revised danger level is Lv-1, the warning zone, and the collision avoidance strategy is to trigger an audible and visual alarm to alert the driver of an obstacle ahead, while maintaining the speed; the revised danger level is Lv-2, the deceleration zone, and the collision avoidance strategy is to issue a speed command, causing the vehicle to reduce its speed to a low level; the revised danger level is Lv-3, the emergency stop zone, and the collision avoidance strategy is to issue an emergency braking signal, causing the vehicle to stop immediately.

[0185] Step S2058: Based on the collision avoidance strategy, control the engineering machinery to perform collision avoidance processing.

[0186] Specifically, electronic devices can send collision avoidance strategies to the motion control unit of the target construction machinery, and complete closed-loop collision avoidance control based on the motion control unit.

[0187] The collision avoidance method for engineering machinery provided in this application determines the ground region of interest (ROI) corresponding to the lidar based on the initial point cloud data for each outrigger, the installation height and elevation angle of the lidar corresponding to the outrigger, and the geometric parameters of the engineering machinery. The ground acquisition space range of the lidar is delineated by combining the actual installation parameters of the equipment and the geometric features of the machine body, precisely limiting the search area for ground points; interference from non-ground space-irrelevant point clouds is eliminated, narrowing the data range for subsequent calculations, reducing computational overhead, and improving the accuracy of ground candidate point extraction. Based on the ground ROI, a subset of ground candidate points falling within the ground ROI is selected from the initial point cloud data. Potential ground points are extracted from the original point cloud, and non-ground points such as structures and obstacles are removed, with quality assessment only performed on ground-related points; this avoids interference from non-ground target points in the point cloud quality evaluation, ensuring that quality detection focuses only on the true state of the ground echo. The number of effective ground points corresponding to the ground candidate point subset, the continuity of the ground point cloud, and the mean and variance of the echo intensity distribution are calculated. Quantitative evaluation indicators are extracted from multiple dimensions, including the number of points, spatial geometry, and echo reflection characteristics, moving away from relying on a single indicator to judge the quality of the point cloud. These multi-dimensional features can comprehensively characterize the point cloud degradation caused by water accumulation and specular reflection, providing sufficient quantitative basis for subsequent quality judgment. If the number of valid ground points exceeds a preset threshold, the continuity of the ground point cloud exceeds a preset threshold, the mean distribution exceeds a preset threshold, and the variance is less than a preset threshold, then the initial point cloud data corresponding to the outrigger meets the preset quality conditions. A multi-condition AND joint judgment logic is adopted; only when multiple indicators simultaneously meet the criteria is the point cloud quality deemed acceptable. This avoids misjudgments caused by fluctuations in a single indicator, reliably distinguishing between normal ground echoes and point cloud failures caused by water accumulation and specular reflection, accurately identifying the usability of the ground point cloud for each outrigger radar, and providing reliable judgment results for subsequent graded selection of elevation angle data sources.

[0188] If the initial point cloud data corresponding to one outrigger of the target construction machinery does not meet the preset quality conditions, the ground elevation angle corresponding to the target construction machinery is calculated based on the initial point cloud data corresponding to the other outrigger. This fully utilizes the hardware redundancy resources of the RTG multi-outrigger radar. When the ground point cloud fails due to reflection from water accumulation on one outrigger, the point cloud data from the intact outrigger on the opposite side is directly reused to calculate the ground elevation angle. There is no need to directly switch to the indirect fitting mode for structures; the measured data directly observed by the radar is prioritized to preserve the angle calculation accuracy to the greatest extent possible, avoiding premature use of structure fitting that could lead to loss of detection accuracy. This achieves smooth fault tolerance and degradation under mild water accumulation conditions. When only one radar fails, there is no need to switch to the fallback structure calculation mode; reliable ground elevation angles can be obtained using other radars on the machine. This solves the problem of single-sided radar failure due to water accumulation while ensuring the detection performance of the collision avoidance system. This provides an intermediate layer for the entire graded fault tolerance mechanism. A progressive logic of "local radar available → outboard radar → structure as backup" is formed, which avoids degrading to a lower-precision backup solution when a local radar is interfered with, thus balancing operational accuracy and anti-interference capability. The original elevation angle calculation algorithm remains unchanged. The elevation angle output by the outboard radar has the same physical meaning as the elevation angle output by the local radar. Subsequent equivalent ground plane, obstacle segmentation, and collision avoidance processing can reuse the original algorithm flow, reducing algorithm modifications and improving system engineering feasibility.

[0189] If the initial point cloud data for each outrigger of the target construction machinery does not meet the preset quality conditions, the initial point cloud data for each outrigger is uniformly transformed to the global coordinate system of the target construction machinery. The initial point cloud data after coordinate alignment are then merged to obtain merged initial point cloud data. This eliminates the differences in local coordinate systems of different outrigger lidars, unifies the point clouds of multiple radars to the RTG global coordinate system, achieves spatial alignment of multi-source point clouds, avoids calculation errors caused by coordinate inconsistencies in subsequent filtering and segmentation, and provides a unified data foundation for multi-radar joint processing. The merged initial point cloud data is downsampled to obtain downsampled point cloud data. While preserving the geometric contour features of the object, the total number of point clouds is reduced, reducing the computational overhead of subsequent steps, improving the real-time performance of the algorithm, and meeting the real-time response requirements of the large vehicle collision avoidance system. For each downsampled data point in the downsampled point cloud data, the number of neighboring points in the preset neighborhood area corresponding to the downsampled data point is calculated. The number of neighboring points quantifies the spatial dispersion of the points, providing a quantitative basis for identifying isolated noise points. If the number of neighboring points is less than the preset threshold, the downsampled data point is deleted. The process removes isolated outliers and eliminates scattered invalid points generated by distant floating particles and random scattering, reducing noise interference in subsequent stages. If the number of neighboring points is greater than or equal to a preset threshold, the sampled data point is checked to see if it falls within the ground region of interest (ROI) based on its coordinates. Ground and non-ground points are distinguished, and valid entity points are categorized and branched to prevent ground points from participating in the floating suspicious point identification logic. If a sampled data point falls within the ground ROI, it is deleted. Ground points are directly removed to reduce the amount of data extracted from subsequent structures and to avoid interference from ground points in rigid structure segmentation. If a sampled data point does not fall within the ground ROI, its height data is obtained. Height is introduced as a discrimination criterion to divide the floating space from the space where the structure is located, providing a basis for distinguishing between floating noise and structure entity points. If the height data is less than the preset minimum height for rigid structures, the sampled data point is marked as a floating suspicious point. Points below the minimum height of the structure and in non-ground areas are marked as objects to be checked. Raindrops and specular reflections from accumulated water are mostly concentrated in this area, enabling targeted screening of interference points. If the height data is greater than or equal to the preset minimum height of the rigid structure, the sampled data points are marked as high-confidence protection points for the structure. Points within the structure's height range are protected to prevent them from participating in the noise point removal logic, thus avoiding the misfiltering of valid points from structures such as containers, railings, and columns, ensuring the complete preservation of structure features. For each suspected suspended point, the echo intensity of the single point and the variance of the echo intensity of all sampled data points within the preset neighborhood of the suspected suspended point are calculated.Combining single-point echo intensity and local echo intensity variance, the signal characteristics of raindrop noise and specular reflection noise from water accumulation are respectively characterized, providing multi-dimensional statistical basis for noise identification. If the single-point echo intensity is less than a preset echo intensity threshold, and / or the echo intensity variance is greater than a preset echo intensity variance threshold, then the floating suspicious point is deleted. The combined echo intensity features filter out floating virtual points caused by raindrop scattering and specular reflection from water accumulation, reducing false alarm sources for subsequent obstacle recognition and attitude calculation. The remaining data points are determined as the filtered point cloud dataset. The output is a high-quality point cloud that has undergone multiple rounds of layered filtering, removing various interference noises in the rainy environment while completely preserving the effective point clouds of rigid structures and real obstacles, providing reliable input for subsequent structure point cloud extraction and attitude calculation.

[0190] Then, based on the prior information of the equipment size and yard layout corresponding to the target engineering machinery, the spatial region of interest (ROI) of the rigid structure is delineated in the global coordinate system. Prior knowledge of the yard is used to limit the target search space, filtering out invalid point clouds in ground areas and distant scenes, thus narrowing the range of data to be processed; reducing interference from irrelevant point clouds, lowering the computational load of the algorithm, and improving the computational efficiency of structure extraction. Candidate rigid structure point clouds falling into the ROI are extracted from the filtered point cloud dataset. Potential structure points within the spatial range are quickly screened, completing the initial coarse extraction; irrelevant points not within the ROI are directly excluded, reducing the data processing pressure of subsequent screening and clustering stages. Geometric shape and echo intensity are used to screen the candidate rigid structure point clouds, and constraints are applied based on a preset size library of yard structures to obtain backup rigid structure point clouds. Multi-dimensional joint filtering based on geometric shape, echo reflection characteristics, and actual object size eliminates scattered virtual points and non-structure interference targets; narrowing the candidate set reduces the probability of false point clusters generated in subsequent clustering, improving the confidence of the backup point clouds. Clustering is performed on the point cloud of spare rigid structures to generate at least one independent structure point cluster. Clustering divides the point cloud into independent structure objects based on spatial relationships, achieving separation of different structures and enabling attitude calculations for individual structures, providing a foundation for subsequent multi-structure angle fusion. Each independent structure point cluster is validated to obtain a rigid structure point cloud. Point clusters are validated for point quantity, size, and attitude constraints, eliminating spurious targets formed by noise; deformed and incomplete invalid point clusters are filtered out, outputting a high-reliability rigid structure point cloud. This avoids inputting inferior point clouds into the attitude calculation process, preventing incorrect pitch angles and ensuring the accuracy of ground pitch angle calculations under extreme water accumulation conditions.

[0191] Based on a point cloud of rigid structures, the pitch angle of the rigid structure itself corresponding to the point cloud is calculated. The pitch angle is calculated using the fixed geometric features of the rigid structure, bypassing the invalid ground echo point cloud. Multi-source fusion can also be performed using multiple structure point clouds to reduce attitude calculation errors caused by single-point noise, resulting in a more reliable observed attitude of the structure. The pitch angle of the rigid structure itself is converted to obtain the ground pitch angle corresponding to the target engineering machinery. Utilizing the inherent parallel / perpendicular geometric constraints between the structure and the yard ground, the structure pitch angle is converted into a ground pitch angle, indirectly obtaining the vehicle's pitch attitude relative to the ground. The output ground pitch angle has the same physical meaning as the pitch angle directly observed by radar. Subsequent equivalent ground plane construction, obstacle segmentation, and collision avoidance processes can completely reuse the original algorithm logic without redesigning the entire calculation chain, ensuring the continuous operation of collision avoidance services under extreme water accumulation conditions.

[0192] Finally, the vertical distance from each initial point cloud to the equivalent ground plane is calculated. Using the equivalent ground plane obtained through pitch angle correction as the segmentation benchmark, the height of each point relative to the virtual ground is quantified by the vertical distance, eliminating reliance on invalid local ground point clouds and providing a unified mathematical basis for the division of ground points and obstacle points. If the vertical distance is greater than a preset vertical distance threshold, the initial point cloud is identified as an obstacle candidate point. Points higher than the ground benchmark are quickly separated based on the height threshold, potential obstacle points are screened, and the initial segmentation of ground points and obstacle candidate points is completed, reducing the data scale for subsequent processing. Based on each obstacle candidate point, an obstacle candidate point cloud set is generated. All potential obstacle points are collected to form a unified processing object, facilitating subsequent clustering operations; all point cloud information of suspected targets is centrally retained to avoid the loss of effective obstacle points. The obstacle candidate point cloud set is clustered to generate at least one candidate obstacle cluster block. Spatially adjacent discrete candidate points are aggregated into independent candidate obstacle clusters, transforming the scattered point cloud into object-level candidate objects, realizing the point-to-object conversion and providing a foundation for object dimension verification. Each candidate obstacle cluster is verified to obtain at least one real target obstacle. The validity of point clusters is verified, filtering out false clusters formed by residual noise from water accumulation and reflective virtual points, thus eliminating false alarms; real obstacles are selected to reduce the probability of false alarms in the collision avoidance system. Key feature parameters required for collision avoidance decisions relative to each real target obstacle are calculated; these parameters include the relative distance and position between the target machinery and the real target obstacle, as well as the target size of the real obstacle. Quantitative parameters necessary for collision avoidance decisions are extracted, obtaining not only distance information but also orientation and object size, enabling the determination of whether an obstacle is within the vehicle's driving lane, providing a complete and objective calculation basis for the hierarchical collision avoidance strategy. Based on the key feature parameters required for collision avoidance decisions, a collision avoidance strategy corresponding to the real target obstacle is generated, and the machinery is controlled to perform collision avoidance actions based on this strategy. Matching the corresponding collision avoidance strategy according to the actual danger level of the obstacle, outputting different levels of control strategies such as warning, deceleration, and emergency stop, realizing graded collision avoidance protection for RTG trolleys under extreme working conditions such as rain and water accumulation, and ensuring the safety of automated operation.

[0193] This embodiment also provides an engineering machinery anti-collision device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0194] This embodiment provides an anti-collision device for engineering machinery, applied to the target engineering machinery, such as... Figure 3 As shown, it includes: The acquisition module 301 is used to acquire the initial point cloud data collected by the lidar installed on each leg of the target construction machinery under a preset operation scenario. The determination module 302 is used to perform quality inspection on the initial point cloud data corresponding to each leg and determine whether the initial point cloud data corresponding to each leg meets the preset quality conditions. The first calculation module 303 is used to calculate the ground pitch angle corresponding to the target engineering machinery based on the detection results; The second calculation module 304 is used to calculate the equivalent ground plane based on the ground pitch angle; The processing module 305 is used to perform collision avoidance processing on engineering machinery based on the equivalent ground plane.

[0195] In some optional implementations, the determining module 302 is specifically used to determine the ground region of interest corresponding to the lidar based on the installation height, installation pitch angle, and geometric parameters of the engineering machinery of the initial point cloud data corresponding to each outrigger; based on the ground region of interest, a subset of ground candidate points falling into the ground region of interest is selected from the initial point cloud data; the number of effective ground points, the continuity of the ground point cloud, and the mean and variance of the echo intensity distribution corresponding to the subset of ground candidate points are calculated; if the number of effective ground points is greater than a preset effective point number threshold, the continuity of the ground point cloud is greater than a preset continuity threshold, the mean of the distribution is greater than a preset mean threshold, and the variance of the distribution is less than a preset variance threshold, then the initial point cloud data corresponding to the outrigger is determined to meet the preset quality conditions.

[0196] In some optional implementations, the first calculation module 303 is specifically used to calculate the ground pitch angle of the target construction machinery based on the initial point cloud data corresponding to the other outrigger of the target construction machinery if the initial point cloud data corresponding to one outrigger of the target construction machinery does not meet the preset quality conditions.

[0197] In some optional implementations, the first calculation module 303 is specifically used to extract a rigid structure point cloud set from the initial point cloud data corresponding to each outrigger of the target engineering machinery if the initial point cloud data corresponding to each outrigger does not meet the preset quality conditions; calculate the pitch angle of the rigid structure itself corresponding to the rigid structure point cloud set based on the rigid structure point cloud set; and convert the pitch angle of the rigid structure itself to obtain the ground pitch angle corresponding to the target engineering machinery.

[0198] In some optional implementations, the first calculation module 303 is specifically used to filter the initial point cloud data corresponding to the outriggers to generate a filtered point cloud dataset; delineate the spatial region of interest of the rigid structure in the global coordinate system based on the equipment size and prior information of the yard layout corresponding to the target engineering machinery; extract candidate rigid structure point clouds that fall into the spatial region of interest from the filtered point cloud dataset; perform geometric shape screening and echo intensity screening on the candidate rigid structure point clouds, and constrain them based on the preset size library of yard structures to obtain a backup rigid structure point cloud; perform clustering processing on the backup rigid structure point clouds to generate at least one independent structure point cluster; and verify each independent structure point cluster to obtain a rigid structure point cloud.

[0199] In some optional implementations, the first calculation module 303 is specifically used to uniformly convert the initial point cloud data corresponding to each outrigger to the global coordinate system of the target engineering machinery, merge the coordinate-aligned initial point cloud data to obtain merged initial point cloud data; perform downsampling processing on the merged initial point cloud data to obtain downsampled point cloud data; for each downsampled data point in the downsampled point cloud data, calculate the number of neighboring points in the preset neighborhood area corresponding to the downsampled data point; if the number of neighboring points is less than the preset neighboring point number threshold, delete the downsampled data point; if the number of neighboring points is greater than or equal to the preset neighboring point number threshold, detect whether the downsampled data point falls into the ground region of interest based on the data coordinates corresponding to the downsampled data point; if the downsampled data point falls into the ground region of interest... If a region of interest is identified, the downsampled data point is deleted. If the downsampled data point does not fall within the region of interest on the ground, the height data corresponding to the downsampled data point is obtained. If the height data is less than the preset minimum height of a rigid structure, the downsampled data point is marked as a suspected floating point. If the height data is greater than or equal to the preset minimum height of a rigid structure, the downsampled data point is marked as a high-confidence protection point of the structure. For each suspected floating point, the single-point echo intensity corresponding to the suspected floating point and the echo intensity variance corresponding to all downsampled data points in the preset neighborhood of the suspected floating point are calculated. If the single-point echo intensity is less than the preset echo intensity threshold, and / or the echo intensity variance is greater than the preset echo intensity variance threshold, the suspected floating point is deleted. The remaining data points are determined as the filtered point cloud dataset.

[0200] In some optional implementations, the processing module 305 is specifically used to calculate the vertical distance between each initial point cloud in each initial point cloud data and the equivalent ground plane; if the vertical distance is greater than a preset vertical distance threshold, the initial point cloud is determined as an obstacle candidate point; based on each obstacle candidate point, an obstacle candidate point cloud set is generated; the obstacle candidate point cloud set is clustered to generate at least one candidate obstacle cluster block; each candidate obstacle cluster block is verified to obtain at least one target real obstacle; the key feature parameters required for collision avoidance decision-making of the target engineering machinery relative to each target real obstacle are calculated; the key feature parameters required for collision avoidance decision-making include the relative distance and relative position between the target engineering machinery and the target real obstacle, as well as the target size corresponding to the target real obstacle; based on the key feature parameters required for collision avoidance decision-making, a collision avoidance strategy corresponding to the target real obstacle is generated; based on the collision avoidance strategy, the engineering machinery is controlled to perform collision avoidance processing.

[0201] The engineering machinery anti-collision device provided in this embodiment of the invention can execute the engineering machinery anti-collision method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0202] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0203] The following is a detailed reference. Figure 4 This diagram illustrates a suitable structure for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, a graphics processing unit, etc.) 01, which can perform various appropriate actions and processes based on a program stored in a read-only memory (ROM) 02 or a program loaded from a memory 08 into a random access memory (RAM) 03. The RAM 03 also stores various programs and data required for the operation of the electronic device. The processor 01, ROM 02, and RAM 03 are interconnected via a bus 04. An input / output (I / O) interface 05 is also connected to the bus 04.

[0204] Typically, the following devices can be connected to I / O interface 05: input devices 06 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 07 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 08 including, for example, magnetic tapes, hard disks, etc.; and communication devices 09. Communication device 09 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0205] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 09, or installed from memory 08, or installed from ROM 02. When the computer program is executed by processor 01, it performs the functions defined in the engineering machinery collision avoidance method of the present invention.

[0206] Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0207] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the engineering machinery collision avoidance method shown in the above embodiments is implemented.

[0208] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0209] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for preventing collisions on engineering machinery, characterized in that, Applied to target construction machinery, the method includes: Under a preset operating scenario, the initial point cloud data collected by the lidar installed on each leg of the target engineering machinery is obtained; The initial point cloud data corresponding to each of the legs is subjected to quality detection to determine whether the initial point cloud data corresponding to each of the legs meets the preset quality conditions. Based on the test results, calculate the ground pitch angle corresponding to the target construction machinery; Based on the ground pitch angle, the equivalent ground plane is calculated; Based on the equivalent ground plane, the engineering machinery is subjected to collision protection treatment.

2. The method according to claim 1, characterized in that, The step of performing quality detection on the initial point cloud data corresponding to each of the aforementioned legs to determine whether the initial point cloud data corresponding to each of the aforementioned legs meets preset quality conditions includes: For the initial point cloud data corresponding to each of the outriggers, the ground region of interest corresponding to the lidar is determined based on the installation height, installation pitch angle of the lidar corresponding to the outrigger and the geometric parameters of the engineering machinery. Based on the region of interest, a subset of candidate ground points falling within the region of interest is selected from the initial point cloud data; Calculate the number of valid ground points, the continuity of the ground point cloud, and the mean and variance of the echo intensity distribution corresponding to the ground candidate point subset; If the number of valid ground points is greater than a preset valid point number threshold, the continuity of the ground point cloud is greater than a preset continuity threshold, the mean of the distribution is greater than a preset mean threshold, and the variance of the distribution is less than a preset variance threshold, then the initial point cloud data corresponding to the outrigger is determined to meet the preset quality conditions.

3. The method according to claim 1, characterized in that, The step of calculating the ground pitch angle corresponding to the target construction machinery based on the detection results includes: If the initial point cloud data corresponding to one outrigger of the target construction machinery does not meet the preset quality conditions, then the ground pitch angle corresponding to the target construction machinery is calculated based on the initial point cloud data corresponding to the other outrigger of the target construction machinery.

4. The method according to any one of claims 1 or 3, characterized in that, The step of performing anti-collision treatment on the construction machinery based on the test results includes: If the initial point cloud data corresponding to each of the outriggers of the target engineering machinery does not meet the preset quality conditions, then the rigid structure point cloud set is extracted from the initial point cloud data corresponding to each outrigger. Based on the set of points of the rigid structure, calculate the pitch angle of the rigid structure itself corresponding to the set of points of the rigid structure; The pitch angle of the rigid structure itself is converted to obtain the ground pitch angle corresponding to the target engineering machinery.

5. The method according to claim 4, characterized in that, The step of extracting the rigid structure point cloud set from the initial point cloud data corresponding to each of the legs includes: The initial point cloud data corresponding to each of the legs are filtered to generate a filtered point cloud dataset. Based on the prior information of the equipment size and yard layout of the target engineering machinery, the spatial region of interest of the rigid structure is delineated in the global coordinate system; Extract candidate rigid structure point clouds that fall into the spatial region of interest from the filtered point cloud dataset; The candidate rigid structure point set is screened by geometric shape and echo intensity, and constrained based on the preset size library of yard structures to obtain the backup rigid structure point set. Clustering is performed on the set of standby rigid structure points to generate at least one independent structure point cluster; The point clusters of each independent structure are verified to obtain the point cloud of the rigid structure.

6. The method according to claim 4, characterized in that, The step of filtering the initial point cloud data corresponding to each of the legs to generate a filtered point cloud dataset includes: The initial point cloud data corresponding to each of the outriggers are uniformly converted to the global coordinate system of the target engineering machinery, and the initial point cloud data after coordinate alignment are merged to obtain merged initial point cloud data. The initial point cloud data to be merged is downsampled to obtain downsampled point cloud data; For each downsampled data point in the downsampled point cloud data, calculate the number of neighboring points within the preset neighborhood area corresponding to the downsampled data point; If the number of neighboring points is less than a preset threshold for the number of neighboring points, then the downsampled data points are deleted. If the number of neighboring points is greater than or equal to the preset threshold for the number of neighboring points, then based on the data coordinates corresponding to the downsampled data points, it is detected whether the downsampled data points fall into the region of interest on the ground. If the downsampled data point falls into the region of interest on the ground, then the downsampled data point is deleted; If the downsampled data point does not fall within the region of interest on the ground, then the height data corresponding to the downsampled data point is obtained; If the height data is less than the preset minimum height of the rigid structure, the downsampled data point is marked as a floating suspicious point; If the height data is greater than or equal to the preset minimum height of the rigid structure, then the downsampled data point is marked as a high-confidence protection point of the structure; For each of the aforementioned suspended suspicious points, calculate the single-point echo intensity corresponding to the suspended suspicious point and the variance of the echo intensity corresponding to all downsampled data points in the preset neighborhood of the suspended suspicious point; If the single-point echo intensity is less than a preset echo intensity threshold, and / or the echo intensity variance is greater than a preset echo intensity variance threshold, then the floating suspicious point will be deleted. The retained data points are determined as the filtered point cloud dataset.

7. The method according to claim 1, characterized in that, The collision avoidance treatment of the engineering machinery based on the equivalent ground plane includes: Calculate the vertical distance between each initial point cloud in each of the initial point cloud data and the equivalent ground plane; If the vertical distance is greater than a preset vertical distance threshold, the initial point cloud is determined as an obstacle candidate point; Based on each of the aforementioned obstacle candidate points, an obstacle candidate point cloud set is generated; The candidate obstacle point cloud set is clustered to generate at least one candidate obstacle cluster block; Each candidate obstacle cluster is verified to obtain at least one real target obstacle; Calculate the key feature parameters required for collision avoidance decisions of the target engineering machinery relative to each of the target real obstacles; the key feature parameters required for collision avoidance decisions include the relative distance and relative position between the target engineering machinery and the target real obstacles, as well as the target size corresponding to the target real obstacles; Based on the key feature parameters required for the collision avoidance decision, a collision avoidance strategy corresponding to the target real obstacle is generated; Based on the aforementioned collision avoidance strategy, the engineering machinery is controlled to perform collision avoidance procedures.

8. A collision avoidance device for engineering machinery, characterized in that, Applied to target construction machinery, the device includes: The acquisition module is used to acquire the initial point cloud data collected by the lidar installed on each leg of the target engineering machinery under a preset operation scenario. The determination module is used to perform quality detection on the initial point cloud data corresponding to each of the legs, and determine whether the initial point cloud data corresponding to each of the legs meets the preset quality conditions. The first calculation module is used to calculate the ground pitch angle corresponding to the target engineering machinery based on the detection results; The second calculation module is used to calculate the equivalent ground plane based on the ground pitch angle; The processing module is used to perform anti-collision processing on the engineering machinery based on the equivalent ground plane.

9. An engineering machinery, characterized in that, include: The engineering machinery body and electronic equipment; the electronic equipment includes a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the engineering machinery collision avoidance method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the engineering machinery collision avoidance method according to any one of claims 1 to 7.

11. A computer program product, characterized in that, It includes computer instructions for causing a computer to execute the engineering machinery collision avoidance method according to any one of claims 1 to 7.

Images