Image recognition-based cruising electric vehicle dynamic obstacle avoidance method and system

Abstract

The invention discloses a cruising electric vehicle dynamic obstacle avoidance method and system based on image recognition, and relates to the technical field of vehicle obstacle avoidance, the method comprises the following steps: acquiring cruising characteristic information of a target cruising vehicle, and establishing a standard cruising control area; carrying out multi-angle environment image acquisition, and carrying out object identification on the acquired multi-angle environment image; in combination with the real-time coordinate set and the corresponding real-time cruise action time sequence information, carrying out follow-up correction on the standard cruise control area to obtain a real-time cruise control body, and carrying out collision risk judgment; and controlling the target cruising vehicle to perform dynamic obstacle avoidance according to the collision risk judgment result. The technical problems that an existing electric vehicle is single in obstacle avoidance strategy, response lags behind or excessive obstacle avoidance affects normal cruising efficiency are solved.

Description

Technical Field

[0001] This application relates to the field of vehicle obstacle avoidance technology, specifically to a dynamic obstacle avoidance method and system for cruising electric vehicles based on image recognition. Background Technology

[0002] With the acceleration of urbanization and the development of the sharing economy, electric cruising vehicles, with their environmentally friendly and flexible characteristics, are widely used in scenarios such as short-distance urban shuttles, scenic area tours, and park commuting. Electric cruising vehicles typically operate in complex environments, and their dynamic obstacle avoidance capabilities directly affect operational safety and service quality.

[0003] However, in existing technologies, traditional vehicle obstacle avoidance methods often rely solely on a single camera to collect image information, resulting in blind spots in environmental perception. Furthermore, when constructing a safety control area, static or simple dynamic bounding box models are often used, failing to fully integrate the cruising characteristics of the vehicle itself and the real-time motion of surrounding objects for dynamic adjustments. This leads to low accuracy in collision risk assessment and makes it difficult to meet the refined obstacle avoidance needs of cruising electric vehicles in complex dynamic environments. Problems include a single obstacle avoidance strategy, delayed response, or excessive obstacle avoidance affecting normal cruising efficiency. Summary of the Invention

[0004] This application provides a dynamic obstacle avoidance method and system for cruising electric vehicles based on image recognition, which solves the technical problems of existing electric vehicles having a single obstacle avoidance strategy, delayed response, or excessive obstacle avoidance affecting normal cruising efficiency.

[0005] The technical solution to the above-mentioned technical problems in this application is as follows: In a first aspect, this application provides a dynamic obstacle avoidance method for cruising electric vehicles based on image recognition, the method comprising: Obtain the cruising characteristic information of the target cruising vehicle, and establish a standard cruising control area based on the cruising characteristic information; Multi-angle environmental images are acquired using a distributed image acquisition device, and object recognition is performed on the acquired multi-angle environmental images to obtain a real-time coordinate set, wherein the real-time coordinate set includes the vehicle's corrected coordinates and the coordinates of the listed objects. By combining the real-time coordinate set with the corresponding real-time patrol action timing information, the standard patrol control area is dynamically corrected to obtain the real-time patrol control body, and collision risk is determined based on the real-time patrol control body. Based on the collision risk assessment results, the target patrol vehicle is controlled to dynamically avoid obstacles.

[0006] Secondly, this application provides an image recognition-based dynamic obstacle avoidance system for cruising electric vehicles, including: The information acquisition module is used to acquire the cruising characteristic information of the target cruising vehicle and establish a standard cruising control area based on the cruising characteristic information. The image acquisition module is used to acquire multi-angle environmental images through a distributed image acquisition device, and to perform object recognition on the acquired multi-angle environmental images to obtain a real-time coordinate set, wherein the real-time coordinate set includes the vehicle's corrected coordinates and the coordinates of the listed objects. The area correction module is used to combine the real-time coordinate set with the corresponding real-time patrol action timing information to perform dynamic correction on the standard patrol control area, obtain the real-time patrol control body, and make collision risk judgment based on the real-time patrol control body. The vehicle obstacle avoidance module is used to control the target cruising vehicle to dynamically avoid obstacles based on the collision risk assessment results.

[0007] This application provides one or more technical solutions, which have at least the following technical effects or advantages: This application provides a dynamic obstacle avoidance method and system for cruising electric vehicles based on image recognition. First, it acquires the cruising route and movement information of the target cruising vehicle, capturing the vehicle's trajectory and the movement patterns of its movable parts at different time points. Second, based on cruising characteristic information and the intrinsic design data of the target cruising vehicle, a block model set is constructed and envelope analysis is performed to generate a block-envelope model set. This set is then integrated with the cruising route information and dynamic contour model to evolve the cruising process, forming a standard cruising control area that reflects the vehicle's spatial positioning under standard cruising conditions, making the control area more closely match actual cruising conditions. Third, through a predefined follow-up correction window, combined with real-time coordinate sets and real-time cruising movement timing information, the standard cruising control area is dynamically corrected, upgrading the static area into a real-time cruising control body that reflects the vehicle's current motion state and changes in the surrounding environment, thus determining the collision risk level and improving the accuracy of collision risk assessment. Finally, based on different collision risk assessment results, a graded obstacle avoidance strategy is executed, achieving refinement and dynamism in the obstacle avoidance strategy, effectively balancing obstacle avoidance safety and cruising efficiency.

[0008] Through the above technical solutions, this application solves the problems of delayed obstacle avoidance response or excessive obstacle avoidance caused by blind spots in environmental perception, static control areas, and single obstacle avoidance strategies in the prior art, thereby improving the safe operation capability and cruising service quality of electric vehicles in complex dynamic environments. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating the dynamic obstacle avoidance method for cruising electric vehicles based on image recognition provided in an embodiment of this application. Figure 2 This is a schematic diagram of the structure of the image recognition-based dynamic obstacle avoidance system for cruising electric vehicles provided in an embodiment of this application.

[0011] The components represented by each number in the attached diagram are explained below: Information acquisition module 11, image acquisition module 12, area correction module 13, vehicle obstacle avoidance module 14. Detailed Implementation

[0012] This application provides a dynamic obstacle avoidance method and system for cruising electric vehicles based on image recognition, which addresses the technical problems of existing electric vehicles having a single obstacle avoidance strategy, delayed response, or excessive obstacle avoidance affecting normal cruising efficiency.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] In the description of this application, 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 indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0015] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid unnecessarily obscuring the description of this application. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0016] Example 1, as Figure 1 As shown in the embodiments of this application, a dynamic obstacle avoidance method for cruising electric vehicles based on image recognition is provided, including: S10: Obtain the cruising characteristic information of the target cruising vehicle, and establish a standard cruising control area based on the cruising characteristic information; In this embodiment, firstly, cruising characteristic information is obtained, including cruising route information and cruising action information of the target cruising vehicle. The cruising route information is a set of driving trajectories of the target cruising vehicle in a standard cruising scenario extracted based on historical cruising data, including geographical coordinates, driving speed and turning angle at different time points. The cruising action information is a set of action parameters of movable parts of the target cruising vehicle during the cruising process. Movable parts include, but are not limited to, doors, windows and onboard robotic arms. The action parameters include action type, action amplitude and timing characteristics.

[0017] Secondly, a standard cruise control zone is established based on cruise characteristic information. Based on the intrinsic design data of the target cruise vehicles, the dynamic profile changes of the vehicles under different driving conditions are simulated. Then, envelope analysis is performed on the dynamic profile model set to determine the standard cruise control zone.

[0018] This includes obtaining the cruising characteristic information of the target cruising vehicle, including: Obtain cruising route information of the target environment, wherein the cruising route information has a first time sequence marker; Acquire the cruising action information of the target cruising vehicle, wherein the cruising action information includes the timing action information of at least one movable part on the target cruising vehicle, and the timestamp of the timing action information forms a second timing marker. Based on the first time sequence marker and the second time sequence marker, the cruise route information and the cruise action information are time-aligned to obtain the cruise characteristic information.

[0019] In this embodiment, firstly, cruising characteristic information is acquired. Complete operational data of the target cruising vehicle under standard cruising scenarios over the past three months is retrieved through the vehicle's log system. A Kalman filter algorithm is used to denoise the original trajectory data, removing outliers and generating a smooth driving trajectory curve. Simultaneously, the control command logs of movable parts are parsed to extract the trigger time, duration, and magnitude of each action, constructing an action time sequence database. The cruising route information has a first time sequence marker, meaning that each coordinate point on the driving trajectory is marked with a corresponding timestamp in chronological order.

[0020] Secondly, the cruise vehicle's cruise action information is acquired. Sensor feedback data from movable parts is collected in real time via the vehicle's CAN bus. Combined with the control command log, the action parameters are calibrated to ensure the accuracy of the action amplitude data. The timestamps of the timing action information form a second timing mark, which is also arranged in chronological order.

[0021] Then, based on the dynamic time warping algorithm, the first and second time series markers are aligned, unifying the cruise route information and cruise action information onto the same time axis, matching them in the time dimension, and finally integrating them to form cruise characteristic information containing spatiotemporal correlation features. For example, when the target cruise vehicle travels to a specific road segment, its onboard robotic arm will perform a specific extension action at the corresponding time node. After time alignment, the specific position and driving state of the vehicle when the action occurs can be clearly identified.

[0022] A standard patrol control area is established based on the aforementioned patrol characteristic information, including: Obtain the intrinsic design data of the target cruise vehicle, and based on the intrinsic design data, traverse and establish a digital space model of the target cruise vehicle and several movable parts to obtain a block model set; Combining the parade motion information with the segmented model set, envelope analysis is performed to generate a segmented-envelope model set, wherein the segmented-envelope model set includes a base space model, a movable space model, and an envelope space model; Based on the block-envelope model set, a dynamic contour model of the target cruising vehicle is established, wherein the dynamic contour model has a temporal marker; The cruise route information and the dynamic contour model are integrated to perform a cruise evolution based on time-series markers, thereby generating the standard cruise control area.

[0023] In this embodiment, the intrinsic design data of the target cruise vehicle is first obtained, including the vehicle's 3D design drawings, structural parameters of each movable component, and kinematic constraint equations. A digital space model of the vehicle body and each movable component is built using SolidWorks software. The vehicle body model includes fixed structures such as the body, chassis, and wheels, while the movable component models are configured with corresponding motion joints according to their degrees of freedom, forming a set of modular models composed of multiple sub-models. For example, the onboard robotic arm is divided into four sub-models: base, upper arm, lower arm, and end effector. Each sub-model defines its connection relationships with adjacent components and its rotation / movement range.

[0024] Secondly, envelope analysis is performed by combining the cruising motion information with the segmented model set. For the temporal motion information of each movable part, the spatial motion trajectory of the part throughout the entire motion cycle is simulated based on the segmented model set. Taking the vehicle-mounted robotic arm as an example, the motion path of the robotic arm's end effector in three-dimensional space is calculated using ADAMS dynamic simulation software based on its extension angle, rotation speed, and duration. This path is then spatially enveloped to generate a movable spatial model that completely covers all possible positions of the robotic arm. Simultaneously, the vehicle body serves as the base, and its digital spatial model constitutes the base spatial model. The base spatial model is merged with the movable spatial models of all movable parts through Boolean operations to obtain an envelope spatial model reflecting the maximum spatial occupancy of the vehicle as a whole under static and dynamic motion. These three elements together form the segmented-envelope model set.

[0025] Then, a dynamic contour model of the target cruising vehicle is established based on the block-envelope model set. The temporal markers of the dynamic contour model correspond to the temporal markers in the cruising characteristic information, that is, each time node corresponds to a specific vehicle spatial contour state. By concatenating the block-envelope models at different time nodes in chronological order, a dynamic contour sequence that changes over time is formed. For example, at time t1, the door is closed, the robotic arm is retracted, and the dynamic contour model shows the basic contour of the vehicle body; at time t2, the door is opened 30 degrees, the robotic arm is extended to its maximum extent, and the dynamic contour model includes the extended contours after the door is opened and the robotic arm is extended.

[0026] Finally, the cruise route information and dynamic contour models are integrated to perform a time-stamped cruise evolution. The dynamic contour models are spatiotemporally overlaid along the driving trajectory in the cruise route information; that is, at each time node, the dynamic contour model at that moment is placed at the corresponding geographical coordinates, taking into account the vehicle's driving direction and attitude. By performing a spatial union operation on the dynamic contour models at all time nodes throughout the entire cruise cycle, a three-dimensional region that reflects all possible spatial occupancy of the target cruise vehicle under standard cruise conditions is generated, namely the standard cruise control area. This area not only includes the space traversed by the vehicle's driving trajectory but also the spatial extension of movable parts under different action states.

[0027] S20: Multi-angle environmental images are acquired through a distributed image acquisition device, and object recognition is performed on the acquired multi-angle environmental images to obtain a real-time coordinate set, wherein the real-time coordinate set includes the vehicle correction coordinates and the coordinates of the listed objects. In this embodiment, the distributed image acquisition device consists of high-definition industrial cameras installed at different locations on the target cruising vehicle, including a front-view camera, a rear-view camera, side-view cameras, and a roof-mounted panoramic camera. The cameras have a resolution of at least 20 megapixels and a frame rate of 30fps, enabling them to quickly capture moving objects. Each camera is equipped with an IMU (Inertial Measurement Unit) for real-time acquisition of its own attitude angle and motion acceleration data.

[0028] Secondly, object recognition is performed on the acquired multi-angle environmental images. First, the improved YOLOv5 algorithm is used as the basic detection model. Based on the COCO dataset, object samples specific to the parade scene are added for transfer learning. The category label, bounding box coordinates and confidence score of the recognized object are output simultaneously. The recognition results with low confidence scores are verified a second time. Trajectory prediction is performed by combining the object motion continuity of adjacent frame images to eliminate misidentified samples and obtain the real-time coordinate set.

[0029] Furthermore, the real-time coordinate set includes the vehicle's corrected coordinates and the coordinates of the listed objects. The vehicle's corrected coordinates are obtained by dynamically correcting the real-time position coordinates of the target cruising vehicle based on the intrinsic and extrinsic parameter calibration results of each camera in the distributed image acquisition device and the attitude data collected by the IMU inertial measurement unit. The object coordinates refer to the set of real-time three-dimensional spatial coordinates of dynamic or static objects identified in multi-angle environmental images that may pose a collision risk to the target cruising vehicle. Dynamic objects include pedestrians, other vehicles, cyclists, etc., while static objects include fixed obstacles, temporary construction areas, etc.

[0030] Specifically, step S20 in the method includes: The multi-angle environmental images are acquired by image acquisition devices deployed on the side of the vehicle and the roadside; The multi-angle environmental images are preprocessed to remove background objects from a preset exclusion list; The preprocessed multi-angle environmental image is subjected to dual object recognition using computer vision methods, including recognizing vehicle features to obtain the vehicle's corrected coordinates and recognizing specified objects in a preset list to obtain the coordinates of the list objects. The vehicle-mounted sensors acquire real-time cruise movement timing information of movable parts on the target cruise vehicle, and output the vehicle correction coordinates, the coordinates of the list objects, and the real-time cruise movement timing information as the real-time coordinate set.

[0031] In this embodiment, firstly, multi-angle environmental images are acquired using image acquisition devices deployed on the vehicle side and roadside. The vehicle side image acquisition devices include high-definition cameras installed on the left and right sides of the front bumper, below the rearview mirrors, and in the center of the rear bumper of the target cruising vehicle. Each camera covers a 120-degree horizontal field of view, ensuring 360-degree visual coverage around the vehicle. The roadside image acquisition devices are fixed high-definition cameras pre-deployed at key nodes along the cruising route, such as intersections and areas with high pedestrian traffic. Real-time image data is transmitted to the onboard processing unit via a 5G network, forming a multi-source heterogeneous data complementarity with the vehicle side images. All image acquisition devices undergo time synchronization calibration to ensure that the timestamp error between images from different sources does not exceed 10 milliseconds.

[0032] Secondly, multi-angle environmental images are preprocessed to remove background objects from a pre-defined exclusion list. This exclusion list is constructed based on the characteristics of the parade scene and includes static background elements that do not directly affect obstacle avoidance decisions, such as sky areas, fixed building facades, road markings, and greenbelts. The preprocessing process uses semantic segmentation algorithms to classify the images at the pixel level. For example, using the DeepLabv3+ algorithm, pixel regions belonging to the exclusion list categories are marked as transparent backgrounds, while potential obstacle regions in the foreground are preserved. Simultaneously, the preprocessed images undergo illumination compensation and contrast enhancement, and an adaptive histogram equalization algorithm improves image quality under complex lighting conditions such as backlighting and shadows.

[0033] Furthermore, computer vision methods are combined to perform dual object recognition on the preprocessed multi-angle environmental images.

[0034] Vehicle features are identified to obtain the vehicle's corrected coordinates. By extracting vehicle feature points such as wheel hubs, window outlines, and bumper edges from images captured by side cameras, the two-dimensional image feature points are mapped to three-dimensional space using perspective transformation. Combining the intrinsic and extrinsic parameters of each camera and IMU attitude data, the real-time position deviation of the target cruising vehicle relative to the ground coordinate system is calculated, and the initial positioning coordinates are corrected to obtain more accurate corrected vehicle coordinates.

[0035] Simultaneously, the system identifies designated objects from a pre-defined list to obtain their coordinates. This list includes 12 high-risk object categories, such as pedestrians, non-motorized vehicles, motorized vehicles, and movable obstacles. An improved EfficientDet algorithm is used for multi-scale object detection, with differentiated detection thresholds set for different object categories; for example, the confidence threshold for pedestrian detection is set to 0.75, and for motorized vehicles, it is set to 0.65. For each detected designated object, a stereo vision matching algorithm, such as the SGBM algorithm, is used to calculate its three-dimensional spatial coordinates relative to the vehicle. The coordinate data includes the longitudinal distance, lateral distance, height (three-dimensional components), and velocity vector of the object's center point.

[0036] Finally, the system uses onboard sensors to acquire real-time tracking information of the movable parts on the target patrol vehicle, and outputs the vehicle's corrected coordinates, the coordinates of the listed objects, and the real-time tracking information as a real-time coordinate set. The onboard sensors include angle sensors, displacement sensors, and force feedback sensors installed at the joints of each movable part, which collect motion parameters such as door opening angle, window lifting height, and robotic arm end-effector position in real time, with a sampling frequency of 100Hz.

[0037] Specifically, the real-time cruise action sequence information includes the current action state of each movable part, such as "door is opening," "robotic arm remains extended," remaining action duration, and the trigger condition for the next action. These three types of data are encapsulated into structured data frames with a unified timestamp. The vehicle's correction coordinates use the UTM coordinate system, the list object coordinates use the vehicle coordinate system with the vehicle's center of gravity as the origin, and the real-time cruise action sequence information stores the action parameters in JSON format. Together, these three elements constitute the real-time coordinate set.

[0038] Furthermore, the method further includes acquiring multi-angle environmental images using a distributed image acquisition device, performing object recognition on the acquired multi-angle environmental images, and obtaining a real-time coordinate set, and also includes: Based on the collection timestamp of the coordinates of the list objects, match and obtain the preceding coordinates of the list objects; By combining the coordinates of the listed objects with the coordinates of their predecessors, the motion vector information of the objects is calculated and stored in the real-time coordinate set in association with the coordinates of the listed objects.

[0039] In this embodiment, firstly, based on the timestamp of the coordinates of the listed objects, multiple consecutive coordinate data of the object within one second before the current moment are matched and obtained from the historical data cache as the preceding coordinates of the listed objects. For example, if the timestamp of the current listed object coordinates is t0, the coordinate data of five time points—t0-0.2s, t0-0.4s, t0-0.6s, t0-0.8s, and t0-1.0s—are extracted to form the preceding coordinate sequence. For newly entered objects, if the preceding coordinates are insufficient, the coordinate data of the current frame are used for initialization and filling, and the object is marked as a "newly appeared object".

[0040] Secondly, by combining the coordinates of the listed object with its preceding coordinates, the motion vector information of the object is calculated and determined. The least squares method is used to perform linear fitting on the preceding coordinate sequence and the current coordinates to obtain the linear equation of the object's trajectory in the vehicle coordinate system, and then the motion direction angle and instantaneous velocity magnitude are solved.

[0041] Specifically, let the coordinates of the object at time t be (x(t), y(t)). By performing time difference on the preceding coordinate sequence, the displacement increments Δx and Δy at adjacent times are calculated. Then, based on the sampling time interval Δt = 0.2s, the velocities in the x-direction vx = Δx / Δt and the velocities in the y-direction vy = Δy / Δt are calculated. The magnitude of the motion vector is... The direction angle θ = arctan2(vy, vx). For objects with unstable motion, acceleration calculation is introduced. The x-direction acceleration ax and y-direction acceleration ay are estimated by the second-order difference method to predict the position change trend in the next 0.5 seconds.

[0042] Finally, the calculated object motion vector information, including velocity magnitude, direction angle, acceleration components, and predicted position, is associated with the coordinates of the listed objects and stored in a real-time coordinate set.

[0043] S30: Combining the real-time coordinate set with the corresponding real-time patrol action timing information, the standard patrol control area is dynamically corrected to obtain the real-time patrol control body, and collision risk is determined based on the real-time patrol control body; Specifically, by combining the real-time coordinate set with the corresponding real-time patrol action timing information, the standard patrol control area is dynamically corrected to obtain a real-time patrol control body. Collision risk is then assessed based on the real-time patrol control body. Prior to this, the process includes: Define the duration of the dynamic correction window based on the total system response time; The total system response time is the sum of the perception processing delay, vehicle braking response time, and preset safety redundancy time, determined based on statistical analysis.

[0044] In this embodiment, the duration of the adaptive correction window is first defined based on the total system response time. The total system response time is specifically the sum of the perception processing delay, the vehicle braking response time, and the preset safety redundancy time.

[0045] Among them, the perception processing delay refers to the time interval from acquiring environmental images from the distributed image acquisition device to outputting the real-time coordinate set. Through statistical analysis of multiple experimental data, its mean is about 150 milliseconds and the standard deviation is 20 milliseconds. The vehicle braking response time includes the mechanical response delay and hydraulic set-off time of the braking system. According to the braking system parameters of the target cruise vehicle, its cold response time is measured to be 200 milliseconds and its hot response time is 180 milliseconds. The maximum value of the two, 200 milliseconds, is taken as the benchmark. The preset safety redundancy time takes into account the uncertainties in complex environments, such as the appearance of sudden obstacles and sensor measurement errors. According to industry safety standards, it is set to 300 milliseconds.

[0046] Secondly, adding the above three times together, we get the total system response time T = 150ms + 200ms + 300ms = 650ms. Therefore, the duration of the follow-up correction window is defined as 650 milliseconds. That is, when making real-time cruise control corrections, it is necessary to consider the movement state and spatial occupancy changes of the vehicle's movable parts within the next 650 milliseconds to ensure sufficient time for collision risk assessment and obstacle avoidance decisions.

[0047] Furthermore, after defining the time length of the follow-up correction window, the standard cruise control area is dynamically corrected by combining the real-time coordinate set with the corresponding real-time cruise action timing information to obtain the real-time cruise control body and perform collision risk assessment.

[0048] Furthermore, by combining the real-time coordinate set with the corresponding real-time patrol action timing information, the standard patrol control area is dynamically corrected to obtain a real-time patrol control body, including: Based on the real-time cruise action timing information, the real-time dynamic outline model of the target cruise vehicle is determined by traversing the quasi-cruise control area. By combining the predefined follow-up correction window with the real-time dynamic contour model, the standard cruise control area is extracted into intervals, and the standard cruise control body is generated by combining the preset clearance distance constraints and the interval extraction results. Based on the self-corrected coordinates, the standard cruise control body is dynamically corrected, wherein the dynamic correction includes position correction and route correction, to generate the real-time cruise control body.

[0049] In this embodiment, firstly, based on real-time cruise action timing information, the real-time dynamic contour model of the target cruise vehicle is determined by traversing the standard cruise control area. Specifically, the real-time cruise action timing information includes the current action state, action amplitude, and action cycle of each movable component, such as the opening angle of the door, the extension length and rotation angle of the robotic arm, and the lifting height of the window. By comparing the real-time action parameters with the baseline action parameters of the movable components when the standard cruise control area is constructed, the spatial position offset of each component is calculated.

[0050] Secondly, by combining a predefined follow-up correction window with a real-time dynamic contour model, intervals are extracted from the standard cruise control area. Based on preset clearance distance constraints and the interval extraction results, a standard cruise control body is generated. The follow-up correction window has a duration of 650 milliseconds, therefore, it is necessary to predict the motion sequence of each movable component within the next 650 milliseconds. Based on the motion velocity and acceleration parameters in the real-time cruise motion timing information, a linear interpolation method is used to predict the motion state of movable components within each future time sub-window, such as a sub-window every 50 milliseconds, thereby generating the evolution sequence of the real-time dynamic contour model within the next 650 milliseconds.

[0051] Furthermore, interval extraction is performed on the standard cruise control area, specifically capturing the spatial sub-regions corresponding to the evolution sequence of the real-time dynamic contour model within a 650-millisecond time window. Simultaneously, preset clearance distance constraints are applied, dynamically adjusted based on object type and motion state. For example, a minimum clearance distance of 1.5 meters is set for pedestrians, 2.0 meters for other vehicles, and 0.8 meters for static obstacles. The interval extraction results are then fused with the clearance distance constraints, and the extracted spatial sub-regions are expanded to generate a standard cruise control volume incorporating the time dimension.

[0052] Finally, based on the self-corrected coordinates, the standard cruise control body is dynamically corrected. This dynamic correction includes position correction and route correction, generating a real-time cruise control body. Position correction involves spatially translating the standard cruise control body based on the self-corrected coordinates. The self-corrected coordinates reflect the real-time position of the target cruise vehicle in the ground coordinate system. Aligning the center coordinates of the standard cruise control body with the self-corrected coordinates allows the control body to accurately follow the actual position of the vehicle. Route correction considers changes in the vehicle's direction of travel and attitude.

[0053] The vehicle's real-time heading angle is calculated based on the self-corrected coordinate sequence, such as the direction of the line connecting two consecutive self-corrected coordinates. Combined with roll and pitch angle data collected by the IMU (Inertial Measurement Unit), a three-dimensional rotation transformation is performed on the standard cruise control body to match the vehicle's current driving attitude. Through the combined effect of position and route correction, the standard cruise control body is dynamically adjusted to a real-time cruise control body that is completely synchronized with the current motion state of the target cruise vehicle.

[0054] Furthermore, collision risk assessment is performed based on the aforementioned real-time patrol control system, including: Obtain the object motion vector information corresponding to the coordinates of the objects in the list; By combining the coordinates of the objects in the list, the motion vector information of the objects, and the predefined follow-up correction window, the predicted temporal trajectory information of the objects in the list is constructed; Based on the predicted time-series trajectory information and the real-time patrol control body, a spatiotemporal intersection calculation is performed; If the spatiotemporal intersection operation results in an intersection, the collision reaction time is calculated accordingly. Based on the collision reaction time, the object motion vector information, and the preset collision risk discrimination matrix, the collision risk level is determined and output as the collision risk discrimination result. Otherwise, the output collision risk discrimination result is empty.

[0055] In this embodiment, firstly, the motion vector information of the objects corresponding to the coordinates of the listed objects is obtained, including the object's velocity magnitude, direction angle, acceleration components, and predicted position within the next 0.5 seconds. For example, the motion vector information of a pedestrian is a velocity magnitude of 1.2 m / s, a direction angle of 30°, an acceleration of 0.3 m / s² in the x-direction, an acceleration of -0.1 m / s² in the y-direction, and a predicted position of 0.6 meters moved along the direction of motion from the current coordinates.

[0056] Secondly, by combining the coordinates of the listed objects, their motion vector information, and a predefined follow-up correction window, the predicted temporal trajectory information of the listed objects is constructed. A trajectory prediction method based on a kinematic model is adopted, dividing the follow-up correction window into 13 time steps, each step being 50 milliseconds. Based on the object's current position, velocity, and acceleration, the predicted coordinates corresponding to each time step are calculated.

[0057] For objects with non-zero acceleration, the uniform acceleration motion formula is used for calculation: x(t) = x0 + vx × t + 0.5 × ax × t², y(t) = y0 + vy × t + 0.5 × ay × t², where t is the time difference between the current time step and the initial time. For objects with unstable motion, a Markov chain model is introduced to predict the probabilistic direction of motion. Three possible trajectories and their corresponding probability values ​​are generated at each time step, and the trajectory with the highest probability is selected as the primary predicted trajectory.

[0058] Next, a spatiotemporal intersection calculation is performed based on the predicted time-series trajectory information and the real-time patrol control body. The real-time patrol control body is a collection of cubic regions containing time and spatial dimensions, with each time step corresponding to a spatial control body. The predicted coordinates of the listed objects at each time step are geometrically intersected with the corresponding spatial region of the real-time patrol control body. If the predicted coordinates of an object fall within the spatial range of the control body at a certain time step, it is determined that there is a spatiotemporal intersection at that time step.

[0059] Furthermore, the spatiotemporal intersection operation employs an axis-aligned bounding box collision detection algorithm, representing the spatial region of the real-time patrol control body at each time step as a minimum bounding cuboid. The intersection is determined by comparing the three-dimensional components of the object's predicted coordinates with the boundary values ​​of the cuboid.

[0060] Finally, if the spatiotemporal intersection operation results in an intersection, the collision reaction time is calculated accordingly. Based on the collision reaction time, object motion vector information, and a preset collision risk discrimination matrix, the collision risk level is determined and output as the collision risk discrimination result. Otherwise, the output collision risk discrimination result is empty. The collision reaction time is specifically the difference between the time step corresponding to the first detection of the spatiotemporal intersection and the current system time.

[0061] S40: Control the target patrol vehicle to perform dynamic obstacle avoidance based on the collision risk assessment results.

[0062] In this embodiment of the application, the target cruise vehicle is controlled to perform dynamic obstacle avoidance based on the collision risk judgment result. First, the collision risk judgment result output by S30 is received, which includes the corresponding time step, object information risk level.

[0063] The risk levels of objects range from level 1 to level 5, with levels 4 and 5 considered high-risk, level 3 medium-risk, and levels 1 and 2 low-risk. Based on the risk level, the corresponding obstacle avoidance strategy library is triggered.

[0064] Specifically, step S40 in the method includes: If the collision risk assessment result is not empty, the target cruising vehicle is controlled to perform dynamic obstacle avoidance according to the graded obstacle avoidance strategy, wherein the graded obstacle avoidance strategy control includes: If the collision risk assessment result corresponds to the preset low-risk category, a warning signal is issued and the target cruise vehicle is controlled to slow down slightly. If the collision risk assessment result corresponds to the preset medium risk range, a warning signal is issued, the target cruise vehicle is controlled to slow down slightly, and the cruise action of the movable parts on the target cruise vehicle that cause the entropy of the standard cruise control area to increase is terminated. If the collision risk assessment result corresponds to a preset high-risk category, a warning signal is issued and the target cruise vehicle is controlled to brake and avoid the obstacle.

[0065] In this embodiment, firstly, if the collision risk assessment result corresponds to a preset low-risk category, a low-frequency warning signal is emitted for 3 seconds via the vehicle's audio system and dashboard indicator lights. Simultaneously, the electric drive system is controlled to output a deceleration of -0.5 m / s², reducing the vehicle's speed by 5%-10%. During this process, the update frequency of the real-time coordinate set is increased to 50 Hz to more intensively monitor changes in the object's motion state, ensuring that the risk does not escalate further. The volume of the warning signal and the flashing frequency of the lights are automatically adjusted according to ambient noise and light intensity; for example, the volume is increased to 80 decibels in noisy road sections, and a high-frequency strobe mode is switched in strong light environments.

[0066] Secondly, when the collision risk assessment result corresponds to the preset medium risk range, in addition to issuing a high-frequency, rapid warning signal, increasing the volume to 90 decibels, and using alternating red and blue flashing lights, the vehicle will also be controlled to decelerate moderately at a speed of -1.0 m / s², reducing the speed by 15%-20%. At the same time, all movable parts on the target patrol vehicle that may cause an increase in entropy in the standard patrol control area will be immediately terminated.

[0067] Finally, if the collision risk assessment result corresponds to the preset high-risk category, the highest level of safety response will be triggered immediately, issuing a continuous alarm signal, such as a volume ≥100 decibels, and activating the braking system for emergency braking to remind surrounding traffic participants to take evasive action.

[0068] In summary, compared with the prior art, this application achieves... A real-time patrol control system integrating temporal and spatial dimensions was constructed, enabling accurate prediction of collision risks during the dynamic movement of patrol vehicles. Specifically, by defining a dynamic correction window based on the total system response time, key time parameters such as vehicle perception processing, braking response, and safety redundancy are incorporated into the dynamic adjustment mechanism of the control system. In summary, the embodiments of this application have at least the following technical effects: This application provides an image recognition-based dynamic obstacle avoidance method for cruising electric vehicles. First, it acquires the cruising route and motion information of the target cruising vehicle, capturing the vehicle's trajectory and the movement patterns of its movable parts at different time points. Second, based on cruising characteristic information and the intrinsic design data of the target cruising vehicle, a block model set is constructed and envelope analysis is performed to generate a block-envelope model set. This set is then integrated with the cruising route information and dynamic contour model to evolve the cruising process, forming a standard cruising control area that reflects the vehicle's spatial positioning under standard cruising conditions, making the control area more closely match actual cruising conditions. Third, through a predefined follow-up correction window, combined with real-time coordinate sets and real-time cruising motion sequence information, the standard cruising control area is dynamically corrected, upgrading the static area into a real-time cruising control body that reflects the vehicle's current motion state and changes in the surrounding environment, thus determining the collision risk level and improving the accuracy of collision risk assessment. Finally, based on different collision risk assessment results, a graded obstacle avoidance strategy is executed, achieving refinement and dynamism in the obstacle avoidance strategy, effectively balancing obstacle avoidance safety and cruising efficiency.

[0069] Through the above technical solutions, this application solves the problems of delayed obstacle avoidance response or excessive obstacle avoidance caused by blind spots in environmental perception, static control areas, and single obstacle avoidance strategies in the prior art, thereby improving the safe operation capability and cruising service quality of electric vehicles in complex dynamic environments.

[0070] Example 2, as Figure 2 As shown, based on the same inventive concept as the image recognition-based dynamic obstacle avoidance method for cruising electric vehicles provided in Embodiment 1, this application also provides an image recognition-based dynamic obstacle avoidance system for cruising electric vehicles, including: The information acquisition module 11 is used to acquire the cruising characteristic information of the target cruising vehicle and establish a standard cruising control area based on the cruising characteristic information. Image acquisition module 12 is used to acquire multi-angle environmental images through a distributed image acquisition device, and to perform object recognition on the acquired multi-angle environmental images to obtain a real-time coordinate set, wherein the real-time coordinate set includes the vehicle correction coordinates and the coordinates of the listed objects. The area correction module 13 is used to combine the real-time coordinate set with the corresponding real-time patrol action timing information to perform follow-up correction on the standard patrol control area, obtain the real-time patrol control body, and perform collision risk judgment based on the real-time patrol control body. The vehicle obstacle avoidance module 14 is used to control the target cruising vehicle to perform dynamic obstacle avoidance based on the collision risk assessment result.

[0071] Furthermore, in one embodiment of the application, obtaining the cruising characteristic information of the target cruising vehicle includes: Obtain cruising route information of the target environment, wherein the cruising route information has a first time sequence marker; Acquire the cruising action information of the target cruising vehicle, wherein the cruising action information includes the timing action information of at least one movable part on the target cruising vehicle, and the timestamp of the timing action information forms a second timing marker. Based on the first time sequence marker and the second time sequence marker, the cruise route information and the cruise action information are time-aligned to obtain the cruise characteristic information.

[0072] Furthermore, in one embodiment of the application, establishing a standard cruise control area based on the cruise characteristic information includes: Obtain the intrinsic design data of the target cruise vehicle, and based on the intrinsic design data, traverse and establish a digital space model of the target cruise vehicle and several movable parts to obtain a block model set; Combining the parade motion information with the segmented model set, envelope analysis is performed to generate a segmented-envelope model set, wherein the segmented-envelope model set includes a base space model, a movable space model, and an envelope space model; Based on the block-envelope model set, a dynamic contour model of the target cruising vehicle is established, wherein the dynamic contour model has a temporal marker; The cruise route information and the dynamic contour model are integrated to perform a cruise evolution based on time-series markers, thereby generating the standard cruise control area.

[0073] In one embodiment, the image acquisition module 12 is specifically used for: The multi-angle environmental images are acquired by image acquisition devices deployed on the side of the vehicle and the roadside; The multi-angle environmental images are preprocessed to remove background objects from a preset exclusion list; The preprocessed multi-angle environmental image is subjected to dual object recognition using computer vision methods, including recognizing vehicle features to obtain the vehicle's corrected coordinates and recognizing specified objects in a preset list to obtain the coordinates of the list objects. The vehicle-mounted sensors acquire real-time cruise movement timing information of movable parts on the target cruise vehicle, and output the vehicle correction coordinates, the coordinates of the list objects, and the real-time cruise movement timing information as the real-time coordinate set.

[0074] Furthermore, in one embodiment, the method of acquiring multi-angle environmental images using a distributed image acquisition device, and performing object recognition on the acquired multi-angle environmental images to obtain a real-time coordinate set, further includes: Based on the collection timestamp of the coordinates of the list objects, match and obtain the preceding coordinates of the list objects; By combining the coordinates of the listed objects with the coordinates of their predecessors, the motion vector information of the objects is calculated and stored in the real-time coordinate set in association with the coordinates of the listed objects.

[0075] Furthermore, by combining the real-time coordinate set with the corresponding real-time patrol action timing information, the standard patrol control area is dynamically corrected to obtain a real-time patrol control volume, and collision risk is determined based on the real-time patrol control volume. Prior to this, the process includes: Define the duration of the dynamic correction window based on the total system response time; The total system response time is the sum of the perception processing delay, vehicle braking response time, and preset safety redundancy time, determined based on statistical analysis.

[0076] By combining the real-time coordinate set with the corresponding real-time patrol action timing information, the standard patrol control area is dynamically corrected to obtain a real-time patrol control body, including: Based on the real-time cruise action timing information, the real-time dynamic outline model of the target cruise vehicle is determined by traversing the quasi-cruise control area. By combining the predefined follow-up correction window with the real-time dynamic contour model, the standard cruise control area is extracted into intervals, and the standard cruise control body is generated by combining the preset clearance distance constraints and the interval extraction results. Based on the self-corrected coordinates, the standard cruise control body is dynamically corrected, wherein the dynamic correction includes position correction and route correction, to generate the real-time cruise control body.

[0077] Furthermore, collision risk assessment is performed based on the aforementioned real-time patrol control system, including: Obtain the object motion vector information corresponding to the coordinates of the objects in the list; By combining the coordinates of the objects in the list, the motion vector information of the objects, and the predefined follow-up correction window, the predicted temporal trajectory information of the objects in the list is constructed; Based on the predicted time-series trajectory information and the real-time patrol control body, a spatiotemporal intersection calculation is performed; If the spatiotemporal intersection operation results in an intersection, the collision reaction time is calculated accordingly. Based on the collision reaction time, the object motion vector information, and the preset collision risk discrimination matrix, the collision risk level is determined and output as the collision risk discrimination result. Otherwise, the output collision risk discrimination result is empty.

[0078] In one embodiment, the vehicle obstacle avoidance module 14 is specifically used for: If the collision risk assessment result is not empty, the target cruising vehicle is controlled to perform dynamic obstacle avoidance according to the graded obstacle avoidance strategy, wherein the graded obstacle avoidance strategy control includes: If the collision risk assessment result corresponds to the preset low-risk category, a warning signal is issued and the target cruise vehicle is controlled to slow down slightly. If the collision risk assessment result corresponds to the preset medium risk range, a warning signal is issued, the target cruise vehicle is controlled to slow down slightly, and the cruise action of the movable parts on the target cruise vehicle that cause the entropy of the standard cruise control area to increase is terminated. If the collision risk assessment result corresponds to a preset high-risk category, a warning signal is issued and the target cruise vehicle is controlled to brake and avoid the obstacle.

[0079] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0080] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0081] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A dynamic obstacle avoidance method for cruising electric vehicles based on image recognition, characterized in that, include: Obtain the cruising characteristic information of the target cruising vehicle, and establish a standard cruising control area based on the cruising characteristic information; Multi-angle environmental images are acquired using a distributed image acquisition device, and object recognition is performed on the acquired multi-angle environmental images to obtain a real-time coordinate set, wherein the real-time coordinate set includes the vehicle's corrected coordinates and the coordinates of the listed objects. By combining the real-time coordinate set with the corresponding real-time patrol action timing information, the standard patrol control area is dynamically corrected to obtain the real-time patrol control body, and collision risk is determined based on the real-time patrol control body. Based on the collision risk assessment results, the target patrol vehicle is controlled to dynamically avoid obstacles.

2. The image recognition-based dynamic obstacle avoidance method for cruising electric vehicles as described in claim 1, characterized in that, Obtain cruising characteristic information of the target cruising vehicle, including: Obtain cruising route information of the target environment, wherein the cruising route information has a first time sequence marker; Acquire the cruising action information of the target cruising vehicle, wherein the cruising action information includes the timing action information of at least one movable part on the target cruising vehicle, and the timestamp of the timing action information forms a second timing marker. Based on the first time sequence marker and the second time sequence marker, the cruise route information and the cruise action information are time-aligned to obtain the cruise characteristic information.

3. The image recognition-based dynamic obstacle avoidance method for cruising electric vehicles as described in claim 2, characterized in that, A standard patrol control area is established based on the aforementioned patrol characteristic information, including: Obtain the intrinsic design data of the target cruise vehicle, and based on the intrinsic design data, traverse and establish a digital space model of the target cruise vehicle and several movable parts to obtain a block model set; Combining the parade motion information with the segmented model set, envelope analysis is performed to generate a segmented-envelope model set, wherein the segmented-envelope model set includes a base space model, a movable space model, and an envelope space model; Based on the block-envelope model set, a dynamic contour model of the target cruising vehicle is established, wherein the dynamic contour model has a temporal marker; The cruise route information and the dynamic contour model are integrated to perform a cruise evolution based on time-series markers, thereby generating the standard cruise control area.

4. The image recognition-based dynamic obstacle avoidance method for cruising electric vehicles as described in claim 1, characterized in that, Multi-angle environmental image acquisition is performed using a distributed image acquisition device, and object recognition is performed on the acquired multi-angle environmental images to obtain a real-time coordinate set, including: The multi-angle environmental images are acquired by image acquisition devices deployed on the side of the vehicle and the roadside; The multi-angle environmental images are preprocessed to remove background objects from a preset exclusion list; The preprocessed multi-angle environmental image is subjected to dual object recognition using computer vision methods, including recognizing vehicle features to obtain the vehicle's corrected coordinates and recognizing specified objects in a preset list to obtain the coordinates of the list objects. The vehicle-mounted sensors acquire real-time cruise movement timing information of movable parts on the target cruise vehicle, and output the vehicle correction coordinates, the coordinates of the list objects, and the real-time cruise movement timing information as the real-time coordinate set.

5. The image recognition-based dynamic obstacle avoidance method for cruising electric vehicles as described in claim 1, characterized in that, The system also includes acquiring multi-angle environmental images using a distributed image acquisition device, performing object recognition on the acquired multi-angle environmental images, and obtaining a real-time coordinate set. Based on the collection timestamp of the coordinates of the list objects, match and obtain the preceding coordinates of the list objects; By combining the coordinates of the listed objects with the coordinates of their predecessors, the motion vector information of the objects is calculated and stored in the real-time coordinate set in association with the coordinates of the listed objects.

6. The image recognition-based dynamic obstacle avoidance method for cruising electric vehicles as described in claim 1, characterized in that, Combining the real-time coordinate set with the corresponding real-time patrol action timing information, the standard patrol control area is dynamically corrected to obtain a real-time patrol control volume. Collision risk is then assessed based on the real-time patrol control volume. Prior to this, the process includes: Define the duration of the dynamic correction window based on the total system response time; The total system response time is the sum of the perception processing delay, vehicle braking response time, and preset safety redundancy time, determined based on statistical analysis.

7. The image recognition-based dynamic obstacle avoidance method for cruising electric vehicles as described in claim 1, characterized in that, By combining the real-time coordinate set with the corresponding real-time patrol action timing information, the standard patrol control area is dynamically corrected to obtain a real-time patrol control body, including: Based on the real-time cruise action timing information, the real-time dynamic outline model of the target cruise vehicle is determined by traversing the quasi-cruise control area. By combining the predefined follow-up correction window with the real-time dynamic contour model, the standard cruise control area is extracted into intervals, and the standard cruise control body is generated by combining the preset clearance distance constraints and the interval extraction results. Based on the self-corrected coordinates, the standard cruise control body is dynamically corrected, wherein the dynamic correction includes position correction and route correction, to generate the real-time cruise control body.

8. The image recognition-based dynamic obstacle avoidance method for cruising electric vehicles as described in claim 1, characterized in that, Collision risk assessment is performed based on the real-time patrol control system, including: Obtain the object motion vector information corresponding to the coordinates of the objects in the list; By combining the coordinates of the objects in the list, the motion vector information of the objects, and the predefined follow-up correction window, the predicted temporal trajectory information of the objects in the list is constructed; Based on the predicted time-series trajectory information and the real-time patrol control body, a spatiotemporal intersection calculation is performed; If the spatiotemporal intersection operation results in an intersection, the collision reaction time is calculated accordingly. Based on the collision reaction time, the object motion vector information, and the preset collision risk discrimination matrix, the collision risk level is determined and output as the collision risk discrimination result. Otherwise, the output collision risk discrimination result is empty.

9. The image recognition-based dynamic obstacle avoidance method for cruising electric vehicles as described in claim 1, characterized in that, Based on the collision risk assessment results, the target patrol vehicle is controlled to dynamically avoid obstacles, including: If the collision risk assessment result is not empty, the target cruising vehicle is controlled to perform dynamic obstacle avoidance according to the graded obstacle avoidance strategy, wherein the graded obstacle avoidance strategy control includes: If the collision risk assessment result corresponds to the preset low-risk category, a warning signal is issued and the target cruise vehicle is controlled to slow down slightly. If the collision risk assessment result corresponds to the preset medium risk range, a warning signal is issued, the target cruise vehicle is controlled to slow down slightly, and the cruise action of the movable parts on the target cruise vehicle that cause the entropy of the standard cruise control area to increase is terminated. If the collision risk assessment result corresponds to a preset high-risk category, a warning signal is issued and the target cruise vehicle is controlled to brake and avoid the obstacle.

10. A dynamic obstacle avoidance system for cruising electric vehicles based on image recognition, characterized in that, The method for performing the image recognition-based dynamic obstacle avoidance method for cruising electric vehicles according to any one of claims 1-9 includes: The information acquisition module is used to acquire the cruising characteristic information of the target cruising vehicle and establish a standard cruising control area based on the cruising characteristic information. The image acquisition module is used to acquire multi-angle environmental images through a distributed image acquisition device, and to perform object recognition on the acquired multi-angle environmental images to obtain a real-time coordinate set, wherein the real-time coordinate set includes the vehicle's corrected coordinates and the coordinates of the listed objects. The area correction module is used to combine the real-time coordinate set with the corresponding real-time patrol action timing information to perform dynamic correction on the standard patrol control area, obtain the real-time patrol control body, and make collision risk judgment based on the real-time patrol control body. The vehicle obstacle avoidance module is used to control the target cruising vehicle to dynamically avoid obstacles based on the collision risk assessment results.

Images