A method and system for constructing vehicle-mounted virtualized human and animal model motion

Abstract

The present invention discloses a construction method and system based on the motion of vehicle-mounted virtual human and animal models, which relates to the field of autonomous driving technology, including obtaining autonomous driving scene image information, obtaining human and animal model information based on digitized scene reconstruction according to the autonomous driving scene image information, and obtaining model morphological data based on morphological analysis according to the human and animal model information. The present invention improves target recognition efficiency by selecting human and animal targets, dynamically adjusts the number of nodes required for image recognition through distance adjustment coefficients and node adjustment coefficients, and improves image recognition efficiency while ensuring target recognition accuracy. The human and animal model information provides a data basis for the construction of a human and animal data model, accurately evaluates the motion state of the target through the human and animal data model and the model morphological data, and ensures the stability and reliability of the human and animal model motion construction.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a method and system for constructing vehicle-mounted virtualized human and animal model motions. Background Art

[0002] With the rapid development of intelligent connected vehicle technology, demand for in-vehicle virtualization scenarios is growing, encompassing areas such as autonomous driving testing, in-car entertainment interaction, and intelligent cockpit design. Dynamic simulation of human and animal models is a key technology in these scenarios. Its core goal is to build high-fidelity, real-time virtual traffic participant models on the vehicle platform, further improving autonomous driving technology. For example, dynamic behavior simulation of virtual pedestrians and animals can be used to verify the safety and robustness of autonomous driving algorithms in complex scenarios.

[0003] At present, the construction of in-vehicle virtual human and animal model motion still has the problem of being unable to accurately identify humans and animals in images through autonomous driving scene images, and often directly uses conventional algorithm technology for identification, and cannot adjust the identification standards according to the actual conditions of the image. Human recognition algorithms often need to detect corresponding points to identify humans, but due to the influence of resolution and other factors, the detection time of some points will be greatly extended, reducing the image recognition efficiency, and it is impossible to set appropriate data model construction standards based on the recognition results, affecting the construction efficiency of human and animal model motion. Summary of the Invention

[0004] In order to solve the above technical problems, a method and system for constructing vehicle-mounted virtual human and animal model motion are provided. This technical solution solves the problem raised in the above background technology that it is impossible to accurately identify human and animals in images through autonomous driving scene images, and that conventional algorithm technology is often used directly for identification, and the identification standards cannot be adjusted according to the actual conditions of the image. Human recognition algorithms often need to detect corresponding points to identify humans, but due to the influence of resolution and other factors, the detection time of some points will be greatly extended, reducing the image recognition efficiency, and it is impossible to set appropriate data model construction standards based on the recognition results, affecting the construction efficiency of human and animal model motion.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0006] A method for constructing vehicle-mounted virtualized human and animal model motion, comprising:

[0007] Acquiring autonomous driving scene image information, where the autonomous driving scene image information represents a scene image during autonomous driving;

[0008] Based on the image information of the autonomous driving scene, the human and animal model information is obtained based on the digital scene reconstruction;

[0009] According to the human and animal model information, based on morphological analysis, the model morphological data is obtained;

[0010] Building a data model and a model skeleton based on the human and animal model information to obtain the human and animal data model;

[0011] Analyze autonomous driving scene images based on human and animal data models and model morphology data to obtain human and animal motion information;

[0012] Obtain motion keyframe information based on the motion information of people and animals;

[0013] Using motion keyframe information, motion simulation is performed on the human and animal data models to complete the model motion construction.

[0014] Preferably, the step of obtaining the human and animal model information based on the digitized scene reconstruction according to the autonomous driving scene image information specifically includes:

[0015] Acquiring a reference scene image based on image preprocessing according to the autonomous driving scene image information, wherein the image preprocessing includes image grayscale conversion and image noise reduction processing;

[0016] Based on the YOLO target detection algorithm, the human and animal targets are selected based on the reference scene image to obtain the model bounding box information. The model bounding box is a rectangular frame.

[0017] Based on the OpenPose model detection requirements, obtain target key node information and standard resolution information;

[0018] Based on the reference scene image, obtaining scene image resolution information;

[0019] The ratio of scene image resolution to standard resolution is used as the node adjustment coefficient;

[0020] Based on the scene image resolution information, obtain the standard height information of the model bounding box;

[0021] According to the model bounding box information, obtain the model bounding box height information;

[0022] Obtaining a distance adjustment coefficient according to the model bounding box height information and the model bounding box standard height information;

[0023] According to the distance adjustment coefficient, the node adjustment coefficient and the target key node information, the character and animal model information is obtained.

[0024] Preferably, obtaining the distance adjustment coefficient according to the model bounding box height information and the model bounding box standard height information specifically includes:

[0025] Based on the scene image resolution information, test the selection of human and animal targets, use the YOLO target detection algorithm to select the test target and obtain the test model bounding box information;

[0026] Based on the OpenPose model, identify the key nodes of the test target and obtain the identification information of the key nodes of the test target;

[0027] Based on the identification information of the key nodes of the test target, the height of the test model bounding box is adjusted until all the key nodes of the target are detected;

[0028] On the basis of detecting all the key nodes of the target, gradually reduce the height of the test model bounding box of the test target in the reference scene image until all the key nodes of the target cannot be identified, and use the height of the test model bounding box at this time as the standard height of the model bounding box;

[0029] According to the standard height of the model bounding box, the ratio of the model bounding box height to the standard height of the model bounding box is used as the distance adjustment factor.

[0030] Preferably, the obtaining of the human and animal model information according to the distance adjustment coefficient, the node adjustment coefficient and the target key node information specifically includes:

[0031] Based on the requirements of motion detection of human and animal targets, target key nodes are divided to obtain first target key nodes and second target key nodes, where the first target key nodes include head nodes, shoulder nodes, elbow nodes, wrist nodes, hip nodes, knee nodes, and ankle nodes;

[0032] Classify the second target key node based on the target part of the person or animal to obtain second target key node classification information, wherein the second target key node classification information includes a second node of the head, a second node of the torso, a second node of the upper limbs, and a second node of the lower limbs;

[0033] According to the classification information of the second target key nodes, the number of nodes of each type of the second target key nodes is obtained;

[0034] The ratio of the number of nodes of each type of second target key nodes to the total number of second target key nodes is used as the weight coefficient of the second target key nodes of that type;

[0035] Obtain the required number of the second target key nodes of the model according to the distance adjustment coefficient, the node adjustment coefficient, and the second target key nodes;

[0036] According to the classification information of the second target key nodes, the product of the weight coefficient of each type of the second target key nodes and the required number of the second target key nodes of the model is used as the identification number of the second target key nodes of that type, and the second target key node identification threshold information is obtained;

[0037] Taking the first target key node as the benchmark, identify and connect the benchmark scene images in the model bounding box to obtain the basic character model;

[0038] Based on the basic character model, identify the second target key nodes until the number of identified second target key nodes meets the second target key node identification threshold;

[0039] Based on the identified second target key nodes and the basic character model, the basic character model is connected with nodes to obtain character and animal model information.

[0040] Preferably, the step of constructing a data model and a model skeleton based on the human and animal model information to obtain the human and animal data model specifically includes:

[0041] Based on the data model of people and animals, perform system stress testing to obtain the maximum number of model vertices;

[0042] According to the human animal model information, obtaining the model second target key node information, wherein the model second target key node represents the second target key node in the human animal model;

[0043] Based on the requirements for building a human and animal data model, obtaining joint structure wiring information, wherein the joint structure wiring includes a node ring composed of model nodes;

[0044] According to the joint structure routing information, taking the second target key node as a reference, the second target key node corresponding to each joint structure routing is used as the routing second target key node;

[0045] According to the joint structure wiring information and the model second target key node information, the model second target key node corresponding to each joint structure wiring in the human animal model is obtained;

[0046] According to the joint structure wiring information, the ratio of the model second target key node to the wiring second target key node in each joint structure wiring is used as the wiring node difference coefficient of each joint structure wiring;

[0047] The product of the maximum value of the wiring node difference coefficient and the maximum number of model vertices is taken as the number of model vertices;

[0048] Get the number of model faces according to the number of model vertices;

[0049] Build a basic data model based on the number of model vertices and model faces;

[0050] Based on forward dynamics and inverse dynamics, the model skeleton is constructed with the model vertices in the basic data model as the benchmark;

[0051] Bind the basic data model to the model skeleton to obtain the character and animal data model;

[0052] The number of model faces is:

[0053]

[0054] Where b is the number of model faces and a is the number of model vertices.

[0055] Preferably, obtaining motion key frame generation information based on the motion information of people and animals specifically includes:

[0056] According to the human or animal data model information, based on morphological analysis, obtaining model morphological data, wherein the model morphological data includes the model head-to-body ratio;

[0057] Constructing movement speed standard information according to the model morphology data, wherein the movement speed standard information includes normal walking speed, normal running speed, and fast running speed;

[0058] Analyze autonomous driving scene images based on human and animal data models to obtain motion speed information;

[0059] Comparing the motion speed information with the motion speed standard information to obtain motion state information;

[0060] Based on the motion state information, motion keyframe information is obtained.

[0061] Furthermore, a construction system based on vehicle-mounted virtualized human and animal model motion is proposed to implement the above construction method, including:

[0062] A main control module, wherein the main control module is used to obtain the number of model vertices based on the human and animal model information, obtain the number of model faces based on the number of model vertices, construct a basic data model based on the number of model vertices and the number of model faces, construct a model skeleton based on forward dynamics and inverse dynamics based on the model vertices in the basic data model, bind the basic data model to the model skeleton, obtain the human and animal data model, obtain model morphological data based on morphological analysis based on the human and animal data model information, construct motion speed standard information based on the model morphological data, obtain motion speed information based on the human and animal data model, compare the motion speed information with the motion speed standard information, obtain motion state information, obtain motion key frame information based on the motion state information, and use the motion key frame information to perform motion simulation on the human and animal data model;

[0063] An information acquisition module is configured to acquire autonomous driving scene image information, obtain a reference scene image based on the autonomous driving scene image information and image preprocessing, select human and animal targets based on the reference scene image and the YOLO target detection algorithm, and obtain model bounding box information. Based on OpenPose model detection requirements, the module obtains target key node information and standard resolution information, and obtains scene image resolution information based on the reference scene image.

[0064] An image recognition module is configured to test the selection of human and animal targets based on scene image resolution information, select the test target using the YOLO target detection algorithm, obtain test model bounding box information, obtain the standard height of the model bounding box based on the test model bounding box information, use the ratio of the model bounding box height to the standard height of the model bounding box as a distance adjustment coefficient based on the standard height of the model bounding box, divide the target key nodes based on the requirements of human and animal target motion detection, obtain the first target key node and the second target key node, identify and connect the benchmark scene image in the model bounding box based on the first target key node, obtain a basic human model, identify the second target key node based on the basic human model, perform node connection on the basic human model based on the identified second target key node and the basic human model, and obtain human and animal model information;

[0065] The display module interacts with the main control module and is used to output and display autonomous driving scene image information, human and animal model information, human and animal data models, and motion key frame information.

[0066] Optionally, the main control module specifically includes:

[0067] A control unit, the control unit being configured to obtain model morphological data based on morphological analysis according to the human or animal data model information, construct motion speed standard information according to the model morphological data, obtain motion speed information according to the human or animal data model, compare the motion speed information with the motion speed standard information, obtain motion state information, obtain motion key frame information based on the motion state information, and perform motion simulation on the human or animal data model using the motion key frame information;

[0068] An information receiving unit, which interacts with the information acquisition module and the image recognition module to receive data and transmit it to the data model unit;

[0069] A data model unit is used to obtain the number of model vertices based on the model information of a person or animal, obtain the number of model faces based on the number of model vertices, construct a basic data model based on the number of model vertices and the number of model faces, and based on forward dynamics and inverse dynamics, construct a model skeleton based on the model vertices in the basic data model, bind the basic data model to the model skeleton, and obtain the data model of a person or animal.

[0070] Optionally, the information acquisition module specifically includes:

[0071] a first acquisition unit, configured to acquire autonomous driving scene image information, acquire a reference scene image based on the autonomous driving scene image information and image preprocessing, and select a human or animal target based on the reference scene image and the YOLO target detection algorithm to acquire model bounding box information;

[0072] The second acquisition unit is used to acquire target key node information and standard resolution information based on the OpenPose model detection requirements, and acquire scene image resolution information based on the reference scene image.

[0073] Optionally, the image recognition module specifically includes:

[0074] a distance assessment unit, the distance assessment unit being configured to test the selection of a human or animal target based on the scene image resolution information, select the test target using the YOLO target detection algorithm, obtain test model bounding box information, obtain a model bounding box standard height based on the test model bounding box information, and use a ratio of the model bounding box height to the model bounding box standard height as a distance adjustment coefficient based on the model bounding box standard height;

[0075] An image model unit is used to divide target key nodes based on the requirements of human and animal target motion detection, obtain first target key nodes and second target key nodes, identify and connect the reference scene image in the model boundary box based on the first target key nodes, obtain a basic human model, identify the second target key nodes based on the basic human model, connect the nodes of the basic human model based on the identified second target key nodes and the basic human model, and obtain human and animal model information.

[0076] Compared with the prior art, the present invention has the following beneficial effects:

[0077] The present invention proposes a construction method and system based on the motion of vehicle-mounted virtual human and animal models. By selecting human and animal targets, the efficiency of human and animal target recognition is improved. The number of nodes required for image recognition is dynamically adjusted through the distance adjustment coefficient and the node adjustment coefficient. On the basis of ensuring the accuracy of target recognition, the image recognition efficiency is improved. The human and animal model information provides a data basis for the construction of the human and animal data model. The human and animal data model and model morphological data are used to accurately evaluate the motion state of the target, thereby ensuring the stability and reliability of the human and animal model motion construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 This is a flow chart of a method for constructing vehicle-mounted virtualized human and animal model motions proposed by the present invention;

[0079] Figure 2 This is a flowchart for obtaining information about human and animal models in the present invention;

[0080] Figure 3 This is a flow chart for obtaining the distance adjustment coefficient in the present invention;

[0081] Figure 4 A flowchart for obtaining the human and animal data model in the present invention;

[0082] Figure 5 This is a schematic diagram of the target key nodes of the human and animal data model in the present invention;

[0083] Figure 6 This is a structural block diagram of a construction system based on vehicle-mounted virtual human and animal model motion proposed by the present invention. DETAILED DESCRIPTION

[0084] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0085] Reference Figure 1 - Figure 5 As shown, a method for constructing a vehicle-mounted virtualized character or animal model motion in an embodiment of the present invention includes:

[0086] Acquiring autonomous driving scene image information, where the autonomous driving scene image information represents a scene image during autonomous driving;

[0087] Based on the image information of the autonomous driving scene, the human and animal model information is obtained based on the digital scene reconstruction;

[0088] Specifically, based on the image information of the autonomous driving scene and the digital scene reconstruction, the human and animal model information is obtained, including:

[0089] Acquiring a reference scene image based on image preprocessing according to the autonomous driving scene image information, wherein the image preprocessing includes image grayscale conversion and image noise reduction processing;

[0090] Based on the YOLO target detection algorithm, the human and animal targets are selected based on the reference scene image to obtain the model bounding box information. The model bounding box is a rectangular frame.

[0091] Based on the OpenPose model detection requirements, obtain target key node information and standard resolution information;

[0092] Based on the reference scene image, obtaining scene image resolution information;

[0093] The ratio of scene image resolution to standard resolution is used as the node adjustment coefficient;

[0094] Based on the scene image resolution information, obtain the standard height information of the model bounding box;

[0095] According to the model bounding box information, obtain the model bounding box height information;

[0096] Obtaining a distance adjustment coefficient according to the model bounding box height information and the model bounding box standard height information;

[0097] According to the distance adjustment coefficient, the node adjustment coefficient and the target key node information, the character and animal model information is obtained.

[0098] This solution uses a pre-processing combination of "image grayscale conversion + noise reduction" to effectively filter out light interference (such as backlighting and shadows) and sensor noise in autonomous driving scenes, improving image contrast and clarity, and laying a high-quality data foundation for subsequent object detection. For example, noise reduction reduces the blurring effect of noise such as raindrops and road reflections on human outlines, ensuring that the YOLO algorithm accurately captures object boundaries and avoids misjudgments or missed detections of the model's bounding box due to image quality defects. By quickly selecting rectangular bounding boxes for people and animals, the algorithm meets the real-time requirements of the on-board system (such as 60FPS high-speed detection) and ensures continuous tracking of objects in dynamic scenes. OpenPose extracts key nodes (such as human joints and animal limb joints) and assigns pose information (such as walking and running) to the model, overcoming the limitation of traditional bounding boxes that only describe the object's position but cannot represent motion characteristics. Together, these two methods enable the virtual model to not only locate the target's spatial coordinates in the scene but also simulate realistic motion postures, significantly improving the accuracy of the autonomous driving system's perception training for complex behaviors such as pedestrians waving and animals jumping.

[0099] It is understandable that OpenPose can have up to 25 key nodes. However, in the actual recognition process, due to the influence of the resolution of the captured image, the distance of the person's position, etc., if all key nodes or a fixed number of nodes are to be accurately identified, not only will the recognition time be longer, but the accuracy of the identified nodes will also vary. Therefore, by introducing the "node adjustment coefficient" and the "distance adjustment coefficient", a unified modeling standard across resolutions and scales is constructed, which improves the efficiency of image recognition.

[0100] Specifically, according to the model bounding box height information and the model bounding box standard height information, the distance adjustment coefficient is obtained, which specifically includes:

[0101] Based on the scene image resolution information, test the selection of human and animal targets, use the YOLO target detection algorithm to select the test target and obtain the test model bounding box information;

[0102] Based on the OpenPose model, identify the key nodes of the test target and obtain the identification information of the key nodes of the test target;

[0103] Based on the identification information of the key nodes of the test target, the height of the test model bounding box is adjusted until all the key nodes of the target are detected;

[0104] On the basis of detecting all the key nodes of the target, gradually reduce the height of the test model bounding box of the test target in the reference scene image until all the key nodes of the target cannot be identified, and use the height of the test model bounding box at this time as the standard height of the model bounding box;

[0105] According to the standard height of the model bounding box, the ratio of the model bounding box height to the standard height of the model bounding box is used as the distance adjustment factor.

[0106] This solution establishes a direct correlation between bounding box size and the ability to identify key target nodes (such as human joints and animal limb endpoints) by gradually adjusting the test model's bounding box height until all key nodes are detected. For example, when a pedestrian target is far from a vehicle, its bounding box height in the image is smaller. If this height is lower than the "model bounding box standard height," the OpenPose model may miss details like the wrist and ankle due to insufficient pixel information, resulting in distorted virtual model poses (for example, inability to recognize a pedestrian raising their hand). Therefore, by using the ratio of the model bounding box height to the model bounding box standard height as a distance adjustment factor, we avoid missing pose information due to minimizing the target's distance, providing complete pose data support for subsequent motion trajectory simulation.

[0107] Specifically, according to the distance adjustment coefficient, the node adjustment coefficient and the target key node information, the character and animal model information is obtained, which specifically includes:

[0108] Based on the requirements of motion detection of human and animal targets, target key nodes are divided to obtain first target key nodes and second target key nodes, where the first target key nodes include head nodes, shoulder nodes, elbow nodes, wrist nodes, hip nodes, knee nodes, and ankle nodes;

[0109] Classify the second target key node based on the target part of the person or animal to obtain second target key node classification information, wherein the second target key node classification information includes a second node of the head, a second node of the torso, a second node of the upper limbs, and a second node of the lower limbs;

[0110] According to the classification information of the second target key nodes, the number of nodes of each type of the second target key nodes is obtained;

[0111] The ratio of the number of nodes of each type of second target key nodes to the total number of second target key nodes is used as the weight coefficient of the second target key nodes of that type;

[0112] Obtain the required number of the second target key nodes of the model according to the distance adjustment coefficient, the node adjustment coefficient, and the second target key nodes;

[0113] According to the classification information of the second target key nodes, the product of the weight coefficient of each type of the second target key nodes and the required number of the second target key nodes of the model is used as the identification number of the second target key nodes of that type, and the second target key node identification threshold information is obtained;

[0114] Taking the first target key node as the benchmark, identify and connect the benchmark scene images in the model bounding box to obtain the basic character model;

[0115] Based on the basic character model, identify the second target key nodes until the number of identified second target key nodes meets the second target key node identification threshold;

[0116] Based on the identified second target key nodes and the basic character model, the basic character model is connected with nodes to obtain character and animal model information.

[0117] In this solution, a hierarchical modeling system of "basic skeleton → detailed features" is constructed by dividing the target key nodes into first target key nodes (core key nodes) and second target key nodes (part subdivision nodes). The first key nodes (head, shoulders, elbows, etc.) constitute the basic motion skeleton of the human body / animal, ensuring that the model maintains the physical rationality of the core posture in dynamic scenes (such as the joint linkage logic when walking), avoiding the distortion of the motion trajectory caused by the missing key nodes (such as "elbow-free" robotic arm movement). The second key nodes are then classified by part (head, torso, etc.), and the modeling resources are allocated differentially based on the weight coefficient (the proportion of the number of nodes). For example, the lower limb nodes have a higher weight in walking scenes, and the motion accuracy of the leg joints is prioritized, making the model closer to the real biomechanical characteristics in complex movements (such as running and turning).

[0118] It is understandable that the number of nodes that can be accurately identified in image recognition processes with different accuracies is also very different. Therefore, a basic character model is first constructed by identifying nodes such as the head, shoulders, and elbows, and then each type of second target key node is identified until the identified second target key node of each type exceeds the recognition threshold of the second target key node of that type. The recognition threshold is the product of the weight coefficient of each type of second target key node and the required number of second target key nodes of the model.

[0119] It should be noted that the second target key node classification information includes the second node of the head, the second node of the torso, the second node of the upper limbs, and the second node of the lower limbs, among which:

[0120] Head nodes include nose, left eye, right eye, left ear, and right ear;

[0121] The torso nodes include left shoulder, right shoulder, left hip, right hip, and neck (implicit): the neck position is calculated by the midpoint of the shoulders and hips (non-explicit key points, the OpenPose 18-point version includes a neck node);

[0122] Upper limb nodes include left elbow, left wrist, right elbow, right wrist, fingers (optional, 25 points included): left thumb, left little finger, right thumb, right little finger;

[0123] The lower limb nodes include the left knee, left ankle, right knee, and right ankle.

[0124] According to the human and animal model information, based on morphological analysis, the model morphological data is obtained;

[0125] Building a data model and a model skeleton based on the human and animal model information to obtain the human and animal data model;

[0126] Specifically, the data model and model skeleton are constructed based on the human and animal model information to obtain the human and animal data model, which specifically includes:

[0127] Based on the data model of people and animals, perform system stress testing to obtain the maximum number of model vertices;

[0128] According to the human animal model information, obtaining the model second target key node information, wherein the model second target key node represents the second target key node in the human animal model;

[0129] Based on the requirements for building a human and animal data model, obtaining joint structure wiring information, wherein the joint structure wiring includes a node ring composed of model nodes;

[0130] According to the joint structure routing information, taking the second target key node as a reference, the second target key node corresponding to each joint structure routing is used as the routing second target key node;

[0131] According to the joint structure wiring information and the model second target key node information, the model second target key node corresponding to each joint structure wiring in the human animal model is obtained;

[0132] According to the joint structure wiring information, the ratio of the model second target key node to the wiring second target key node in each joint structure wiring is used as the wiring node difference coefficient of each joint structure wiring;

[0133] The product of the maximum value of the wiring node difference coefficient and the maximum number of model vertices is taken as the number of model vertices;

[0134] Get the number of model faces according to the number of model vertices;

[0135] Build a basic data model based on the number of model vertices and model faces;

[0136] Based on forward dynamics and inverse dynamics, the model skeleton is constructed with the model vertices in the basic data model as the benchmark;

[0137] Bind the basic data model to the model skeleton to obtain the character and animal data model;

[0138] The number of model faces is:

[0139]

[0140] Where b is the number of model faces and a is the number of model vertices.

[0141] In this solution, the maximum number of model vertices is determined through system stress testing, and the vertex scale is dynamically calculated based on the wiring node difference coefficient (the matching degree between the actual nodes of the joint structure and the standard nodes). A "performance-detail" adaptive adjustment mechanism is constructed. Key wiring nodes are screened by the difference coefficient to ensure the synchronization of model surface movement during skeleton drive (such as the wrinkling of the neck skin when the head turns). This solves the "bone and skin separation" problem of traditional models, provides consistent visual and geometric features for multi-sensor fusion testing (such as matching camera and lidar point clouds), and constructs a high-precision data model that conforms to the laws of real biological movement and adapts to real-time computing constraints. It provides a reliable virtual test carrier for the perception and decision-making algorithms of autonomous driving systems.

[0142] Analyze autonomous driving scene images based on human and animal data models and model morphology data to obtain human and animal motion information;

[0143] Obtain motion keyframe information based on the motion information of people and animals;

[0144] Specifically, according to the motion information of people and animals, the motion key frame generation information is obtained, which specifically includes:

[0145] According to the human or animal data model information, based on morphological analysis, obtaining model morphological data, wherein the model morphological data includes the model head-to-body ratio;

[0146] Constructing movement speed standard information according to the model morphology data, wherein the movement speed standard information includes normal walking speed, normal running speed, and fast running speed;

[0147] Analyze autonomous driving scene images based on human and animal data models to obtain motion speed information;

[0148] Comparing the motion speed information with the motion speed standard information to obtain motion state information;

[0149] Based on the motion state information, motion keyframe information is obtained.

[0150] In this solution, a motion speed standard is established through model morphological data (such as head-to-body ratio), so that the motion characteristics of the virtual model conform to the laws of real biomechanics. The motion state (such as walking and sprinting) is generated by comparing the motion speed information with the standard. A closed-loop response mechanism of "environmental perception → motion decision → keyframe generation" is constructed. The precise extraction of motion keyframe information solves the pain point of "fixed keyframe library cannot cover dynamic scenes" in traditional methods.

[0151]

[0152] In this embodiment, the project frame rate is set to 24fps (24 frames per second), and the time axis unit is frame.

[0153] A. Static Movement (Non-Circular, Single-Posture Design)

[0154] Core principle: The foot controller is fixed, and the upper body (waist, shoulders, head) changes slightly dynamically to avoid stiffness. The keyframes are:

[0155]

[0156] Note: Static movements do not need to be strictly looped. You can design a slight sense of breathing or fine-tuning of the center of gravity within 0-24 frames to avoid being rigid throughout the process.

[0157] B. Normal walking motion (24 frames / cycle, alternating feet)

[0158] Core cycle: 24 frames to complete a full stride (left foot step out → right foot step out → return to the starting position), consistent with the walking rhythm of a real person (about 1.2m / s). The key frames are:

[0159]

[0160] The pose at frame 24 is symmetrical to that at frame 0 (the left and right feet are swapped). When exporting the looped animation, you need to check "Cycle Offset".

[0161] C. Normal running action (16 frames / cycle, including the flying phase, faster tempo)

[0162] Core cycle: 16 frames to complete a running cycle (feet in the air → left and right feet landing alternately), consistent with the running rhythm of a real person (about 3m / s). The key frames are:

[0163]

[0164] Using motion keyframe information, motion simulation is performed on the human and animal data models to complete the model motion construction.

[0165] Reference Figure 6 As shown, further, in combination with the above-mentioned method for constructing a vehicle-mounted virtualized human and animal model motion, a system for constructing a vehicle-mounted virtualized human and animal model motion is proposed, comprising:

[0166] A main control module, wherein the main control module is used to obtain the number of model vertices based on the human and animal model information, obtain the number of model faces based on the number of model vertices, construct a basic data model based on the number of model vertices and the number of model faces, construct a model skeleton based on forward dynamics and inverse dynamics based on the model vertices in the basic data model, bind the basic data model to the model skeleton, obtain the human and animal data model, obtain model morphological data based on morphological analysis based on the human and animal data model information, construct motion speed standard information based on the model morphological data, obtain motion speed information based on the human and animal data model, compare the motion speed information with the motion speed standard information, obtain motion state information, obtain motion key frame information based on the motion state information, and use the motion key frame information to perform motion simulation on the human and animal data model;

[0167] An information acquisition module is configured to acquire autonomous driving scene image information, obtain a reference scene image based on the autonomous driving scene image information and image preprocessing, select human and animal targets based on the reference scene image and the YOLO target detection algorithm, and obtain model bounding box information. Based on OpenPose model detection requirements, the module obtains target key node information and standard resolution information, and obtains scene image resolution information based on the reference scene image.

[0168] An image recognition module is configured to test the selection of human and animal targets based on scene image resolution information, select the test target using the YOLO target detection algorithm, obtain test model bounding box information, obtain the standard height of the model bounding box based on the test model bounding box information, use the ratio of the model bounding box height to the standard height of the model bounding box as a distance adjustment coefficient based on the standard height of the model bounding box, divide the target key nodes based on the requirements of human and animal target motion detection, obtain the first target key node and the second target key node, identify and connect the benchmark scene image in the model bounding box based on the first target key node, obtain a basic human model, identify the second target key node based on the basic human model, perform node connection on the basic human model based on the identified second target key node and the basic human model, and obtain human and animal model information;

[0169] The display module interacts with the main control module and is used to output and display autonomous driving scene image information, human and animal model information, human and animal data models, and motion key frame information.

[0170] Main control module, specifically including:

[0171] A control unit, the control unit being configured to obtain model morphological data based on morphological analysis according to the human or animal data model information, construct motion speed standard information according to the model morphological data, obtain motion speed information according to the human or animal data model, compare the motion speed information with the motion speed standard information, obtain motion state information, obtain motion key frame information based on the motion state information, and perform motion simulation on the human or animal data model using the motion key frame information;

[0172] An information receiving unit, which interacts with the information acquisition module and the image recognition module to receive data and transmit it to the data model unit;

[0173] A data model unit is used to obtain the number of model vertices based on the model information of a person or animal, obtain the number of model faces based on the number of model vertices, construct a basic data model based on the number of model vertices and the number of model faces, and based on forward dynamics and inverse dynamics, construct a model skeleton based on the model vertices in the basic data model, bind the basic data model to the model skeleton, and obtain the data model of a person or animal.

[0174] Information acquisition module, specifically including:

[0175] a first acquisition unit, configured to acquire autonomous driving scene image information, acquire a reference scene image based on the autonomous driving scene image information and image preprocessing, and select a human or animal target based on the reference scene image and the YOLO target detection algorithm to acquire model bounding box information;

[0176] The second acquisition unit is used to acquire target key node information and standard resolution information based on the OpenPose model detection requirements, and acquire scene image resolution information based on the reference scene image.

[0177] Image recognition module, specifically including:

[0178] a distance assessment unit, the distance assessment unit being configured to test the selection of a human or animal target based on the scene image resolution information, select the test target using the YOLO target detection algorithm, obtain test model bounding box information, obtain a model bounding box standard height based on the test model bounding box information, and use a ratio of the model bounding box height to the model bounding box standard height as a distance adjustment coefficient based on the model bounding box standard height;

[0179] An image model unit is used to divide target key nodes based on the requirements of human and animal target motion detection, obtain first target key nodes and second target key nodes, identify and connect the reference scene image in the model boundary box based on the first target key nodes, obtain a basic human model, identify the second target key nodes based on the basic human model, connect the nodes of the basic human model based on the identified second target key nodes and the basic human model, and obtain human and animal model information.

[0180] To sum up, the advantages of the present invention are: by selecting human and animal targets and obtaining model bounding box information, the efficiency of human and animal target recognition is improved, which provides a basis for subsequent image recognition; the number of nodes required for image recognition is dynamically adjusted through the distance adjustment coefficient and the node adjustment coefficient, and the image recognition efficiency is improved while ensuring the target recognition accuracy; by constructing a data model and a model skeleton based on the human and animal model information, a data basis is provided for the construction of the human and animal data model, and the efficiency of data model construction is improved; through the human and animal data model and model morphological data, the automatic driving scene image is analyzed, and the motion state of the target is accurately evaluated, so as to accurately select the key frames, and complete the motion simulation of the human and animal data model with the key frames.

[0181] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a vehicle-mounted virtualized character and animal model movement, characterized in that: include: Acquiring autonomous driving scene image information, where the autonomous driving scene image information represents a scene image during autonomous driving; Based on the image information of the autonomous driving scene, the human and animal model information is obtained based on the digital scene reconstruction; According to the human and animal model information, based on morphological analysis, the model morphological data is obtained; Building a data model and a model skeleton based on the human and animal model information to obtain the human and animal data model; Analyze autonomous driving scene images based on human and animal data models and model morphology data to obtain human and animal motion information; Obtain motion keyframe information based on the motion information of people and animals; Use motion keyframe information to simulate the motion of human and animal data models and complete the model motion construction; The method of obtaining the human and animal model information based on the digitized scene reconstruction according to the autonomous driving scene image information specifically includes: Acquiring a reference scene image based on image preprocessing according to the autonomous driving scene image information, wherein the image preprocessing includes image grayscale conversion and image noise reduction processing; Based on the YOLO target detection algorithm, the human and animal targets are selected based on the reference scene image to obtain the model bounding box information. The model bounding box is a rectangular frame. Based on the OpenPose model detection requirements, obtain target key node information and standard resolution information; Based on the reference scene image, obtaining scene image resolution information; The ratio of scene image resolution to standard resolution is used as the node adjustment coefficient; Based on the scene image resolution information, obtain the standard height information of the model bounding box; According to the model bounding box information, obtain the model bounding box height information; Obtaining a distance adjustment coefficient according to the model bounding box height information and the model bounding box standard height information; According to the distance adjustment coefficient, the node adjustment coefficient and the target key node information, the character and animal model information is obtained.

2. A method for constructing a vehicle-mounted virtualized character and animal model movement according to claim 1, characterized in that: The obtaining of the distance adjustment coefficient according to the model bounding box height information and the model bounding box standard height information specifically includes: Based on the scene image resolution information, test the selection of human and animal targets, use the YOLO target detection algorithm to select the test target and obtain the test model bounding box information; Based on the OpenPose model, identify the key nodes of the test target and obtain the identification information of the key nodes of the test target; Based on the identification information of the key nodes of the test target, the height of the test model bounding box is adjusted until all the key nodes of the target are detected; On the basis of detecting all the key nodes of the target, gradually reduce the height of the test model bounding box of the test target in the reference scene image until all the key nodes of the target cannot be identified, and use the height of the test model bounding box at this time as the standard height of the model bounding box; According to the standard height of the model bounding box, the ratio of the model bounding box height to the standard height of the model bounding box is used as the distance adjustment factor.

3. The method for constructing a vehicle-mounted virtualized character and animal model movement according to claim 1, characterized in that: The obtaining of the character and animal model information according to the distance adjustment coefficient, the node adjustment coefficient and the target key node information specifically includes: Based on the requirements of motion detection of human and animal targets, target key nodes are divided to obtain first target key nodes and second target key nodes, where the first target key nodes include head nodes, shoulder nodes, elbow nodes, wrist nodes, hip nodes, knee nodes, and ankle nodes; Classify the second target key node based on the target part of the person or animal to obtain second target key node classification information, wherein the second target key node classification information includes a second node of the head, a second node of the torso, a second node of the upper limbs, and a second node of the lower limbs; According to the classification information of the second target key nodes, the number of nodes of each type of the second target key nodes is obtained; The ratio of the number of nodes of each type of second target key nodes to the total number of second target key nodes is used as the weight coefficient of the second target key nodes of that type; Obtain the required number of the second target key nodes of the model according to the distance adjustment coefficient, the node adjustment coefficient, and the second target key nodes; According to the classification information of the second target key nodes, the product of the weight coefficient of each type of the second target key nodes and the required number of the second target key nodes of the model is used as the identification number of the second target key nodes of that type, and the second target key node identification threshold information is obtained; Taking the first target key node as the benchmark, identify and connect the benchmark scene images in the model bounding box to obtain the basic character model; Based on the basic character model, identify the second target key nodes until the number of identified second target key nodes meets the second target key node identification threshold; Based on the identified second target key nodes and the basic character model, the basic character model is connected with nodes to obtain character and animal model information.

4. The method for constructing a vehicle-mounted virtualized character and animal model movement according to claim 1, characterized in that: The process of constructing a data model and a model skeleton based on the human and animal model information to obtain the human and animal data model specifically includes: Based on the data model of people and animals, perform system stress testing to obtain the maximum number of model vertices; According to the human animal model information, obtaining the model second target key node information, wherein the model second target key node represents the second target key node in the human animal model; Based on the requirements for building a human and animal data model, obtaining joint structure wiring information, wherein the joint structure wiring includes a node ring composed of model nodes; According to the joint structure routing information, taking the second target key node as a reference, the second target key node corresponding to each joint structure routing is used as the routing second target key node; According to the joint structure wiring information and the model second target key node information, the model second target key node corresponding to each joint structure wiring in the human animal model is obtained; According to the joint structure wiring information, the ratio of the model second target key node to the wiring second target key node in each joint structure wiring is used as the wiring node difference coefficient of each joint structure wiring; The product of the maximum value of the wiring node difference coefficient and the maximum number of model vertices is taken as the number of model vertices; Get the number of model faces according to the number of model vertices; Build a basic data model based on the number of model vertices and model faces; Based on forward dynamics and inverse dynamics, the model skeleton is constructed with the model vertices in the basic data model as the benchmark; Bind the basic data model to the model skeleton to obtain the character and animal data model; The number of model faces is: Where b is the number of model faces and a is the number of model vertices.

5. The method for constructing a vehicle-mounted virtualized character and animal model movement according to claim 1, characterized in that: The step of obtaining motion key frame generation information based on the motion information of the person or animal specifically includes: According to the human or animal data model information, based on morphological analysis, obtaining model morphological data, wherein the model morphological data includes the model head-to-body ratio; Constructing movement speed standard information according to the model morphology data, wherein the movement speed standard information includes normal walking speed, normal running speed, and fast running speed; Analyze autonomous driving scene images based on human and animal data models to obtain motion speed information; Comparing the motion speed information with the motion speed standard information to obtain motion state information; Based on the motion state information, motion keyframe information is obtained.

6. A construction system based on vehicle-mounted virtualized character and animal model motion, used to implement the construction method according to any one of claims 1 to 5, characterized in that: include: A main control module, wherein the main control module is used to obtain the number of model vertices based on the human and animal model information, obtain the number of model faces based on the number of model vertices, construct a basic data model based on the number of model vertices and the number of model faces, construct a model skeleton based on forward dynamics and inverse dynamics based on the model vertices in the basic data model, bind the basic data model to the model skeleton, obtain the human and animal data model, obtain model morphological data based on morphological analysis based on the human and animal data model information, construct motion speed standard information based on the model morphological data, obtain motion speed information based on the human and animal data model, compare the motion speed information with the motion speed standard information, obtain motion state information, obtain motion key frame information based on the motion state information, and use the motion key frame information to perform motion simulation on the human and animal data model; An information acquisition module is configured to acquire autonomous driving scene image information, obtain a reference scene image based on the autonomous driving scene image information and image preprocessing, select human and animal targets based on the reference scene image and the YOLO target detection algorithm, and obtain model bounding box information. Based on OpenPose model detection requirements, the module obtains target key node information and standard resolution information, and obtains scene image resolution information based on the reference scene image. An image recognition module is configured to test the selection of human and animal targets based on scene image resolution information, select the test target using the YOLO target detection algorithm, obtain test model bounding box information, obtain the standard height of the model bounding box based on the test model bounding box information, use the ratio of the model bounding box height to the standard height of the model bounding box as a distance adjustment coefficient based on the standard height of the model bounding box, divide the target key nodes based on the requirements of human and animal target motion detection, obtain the first target key node and the second target key node, identify and connect the benchmark scene image in the model bounding box based on the first target key node, obtain a basic human model, identify the second target key node based on the basic human model, perform node connection on the basic human model based on the identified second target key node and the basic human model, and obtain human and animal model information; The display module interacts with the main control module and is used to output and display autonomous driving scene image information, human and animal model information, human and animal data models, and motion key frame information.

7. A system for constructing vehicle-mounted virtualized character and animal model motion according to claim 6, characterized in that: The main control module specifically includes: A control unit, the control unit being configured to obtain model morphological data based on morphological analysis according to the human or animal data model information, construct motion speed standard information according to the model morphological data, obtain motion speed information according to the human or animal data model, compare the motion speed information with the motion speed standard information, obtain motion state information, obtain motion key frame information based on the motion state information, and perform motion simulation on the human or animal data model using the motion key frame information; An information receiving unit, which interacts with the information acquisition module and the image recognition module to receive data and transmit it to the data model unit; A data model unit is used to obtain the number of model vertices based on the model information of a person or animal, obtain the number of model faces based on the number of model vertices, construct a basic data model based on the number of model vertices and the number of model faces, and based on forward dynamics and inverse dynamics, construct a model skeleton based on the model vertices in the basic data model, bind the basic data model to the model skeleton, and obtain the data model of a person or animal.

8. The system for constructing a vehicle-mounted virtualized character and animal model movement according to claim 6, characterized in that: The information acquisition module specifically includes: a first acquisition unit, configured to acquire autonomous driving scene image information, acquire a reference scene image based on the autonomous driving scene image information and image preprocessing, and select a human or animal target based on the reference scene image and the YOLO target detection algorithm to acquire model bounding box information; The second acquisition unit is used to acquire target key node information and standard resolution information based on the OpenPose model detection requirements, and acquire scene image resolution information based on the reference scene image.

9. The system for constructing a vehicle-mounted virtualized character and animal model movement according to claim 6, characterized in that: The image recognition module specifically includes: a distance assessment unit, the distance assessment unit being configured to test the selection of a human or animal target based on the scene image resolution information, select the test target using the YOLO target detection algorithm, obtain test model bounding box information, obtain a model bounding box standard height based on the test model bounding box information, and use a ratio of the model bounding box height to the model bounding box standard height as a distance adjustment coefficient based on the model bounding box standard height; An image model unit is used to divide target key nodes based on the requirements of human and animal target motion detection, obtain first target key nodes and second target key nodes, identify and connect the reference scene image in the model boundary box based on the first target key nodes, obtain a basic human model, identify the second target key nodes based on the basic human model, connect the nodes of the basic human model based on the identified second target key nodes and the basic human model, and obtain human and animal model information.

Images