Obstacle state identification method and device in automatic driving environment of sightseeing park

Abstract

The invention relates to the technical field of obstacle recognition, in particular to an obstacle state recognition method and device in an automatic driving environment of a sightseeing park, and the method comprises the steps: setting corresponding motion models for different target obstacles; environment data and motion parameters are collected in real time through a multi-sensor module carried on a vehicle; performing preprocessing and feature extraction on the environment data to identify the type of a target obstacle; obtaining a prediction path of each obstacle based on the motion model and the motion parameters; and outputting the predicted path to a control system of the vehicle to perform an obstacle avoidance operation. The objective of the invention is to greatly improve the recognition efficiency of vehicles in a sightseeing park and reduce the recognition failure rate while guaranteeing the experience of tourists.

Description

Technical Field

[0001] The present invention relates to the field of obstacle recognition technology, and in particular to a method and device for identifying obstacle states in an automatic driving environment of a tourist park. Background Art

[0002] The autonomous driving environment in a tourist park refers to a set of autonomous driving vehicle operation systems designed and implemented in a specific area. This type of environment usually includes autonomous driving vehicles, intelligent traffic management systems, and related infrastructure.

[0003] There are various obstacles in the tourist park. Fixed obstacles include buildings, sculptures, decorations, trees, vegetation, seats and benches, etc.; dynamic obstacles include other self-driving vehicles, electric sightseeing vehicles, bicycles, tourists moving in the park, and hand-held items, shopping carts or strollers carried by tourists.

[0004] Existing obstacle recognition mainly monitors the relative distance between the autonomous driving vehicle and the obstacle for simple obstacle avoidance. In tourist parks with a large number of movable objects, it is impossible to predict the position of dynamic obstacles in time, making autonomous driving prone to failure. Summary of the invention

[0005] The purpose of the present invention is to provide a method and device for identifying the status of obstacles in an automatic driving environment of a tourist park, aiming to greatly improve the recognition efficiency of vehicles in the tourist park and reduce the recognition failure rate while ensuring the tourist experience.

[0006] To achieve the above-mentioned object, in a first aspect, the present invention provides a method for identifying obstacle states in an automatic driving environment of a tourist park, comprising setting corresponding motion models for different target obstacles;

[0007] Collect environmental data and motion parameters in real time through multi-sensor modules mounted on vehicles;

[0008] Preprocessing and feature extracting the environmental data to identify the target obstacle type;

[0009] Get the predicted path of each obstacle based on the motion model and motion parameters;

[0010] The predicted path is output to the vehicle's control system for obstacle avoidance.

[0011] The target obstacles include adults, minors, vehicles and fixed objects.

[0012] The specific steps of setting corresponding motion models for different target obstacles include:

[0013] Obtain historical speed data of target obstacles;

[0014] Obtain the behavior of the target obstacle based on historical speed data and mark the behavior;

[0015] The motion model is obtained by training the support vector machine model based on the behavior markers and historical speed data.

[0016] The specific steps of collecting environmental data and motion parameters in real time through a multi-sensor module mounted on a vehicle include:

[0017] Use radar to scan the surrounding environment and determine whether the obstacle is within the first preset distance;

[0018] When the obstacle is within the first preset distance, the steerable camera is started to rotate to the corresponding direction to continuously collect image data at the obstacle;

[0019] A laser rangefinder is used to measure the motion parameters of obstacles.

[0020] The specific steps of using radar to scan the surrounding environment and determine whether an obstacle is within the first preset distance include:

[0021] Activate the radar sensor so that it starts to emit electromagnetic waves and receive the reflected signals;

[0022] Evaluate the strength of the reflected signal to determine the distance and angle of the reflection source;

[0023] Calculate the distance to the target by the time delay of the radar echo;

[0024] Use the Doppler effect to estimate the target's angle, i.e. the direction of the target relative to the radar;

[0025] The difference between the detected obstacle distance and the first preset distance is calculated.

[0026] The specific steps of starting the steerable camera to rotate to a corresponding direction to continuously collect image data at the obstacle when the obstacle is within the first preset distance include:

[0027] The control system sends a rotation instruction to the steerable camera, instructing the camera to turn to the direction where the obstacle is located;

[0028] Adjust the position of the camera according to the angle information provided by the radar so that the camera points to the obstacle;

[0029] The camera is pointed at the obstacle and starts to collect images.

[0030] The specific steps of preprocessing and feature extracting the environmental data to identify the target obstacle type include:

[0031] Convert image data to grayscale image;

[0032] Resize grayscale images to a preset size;

[0033] Use the Sobel algorithm to detect edge-prominent object contours in the resized grayscale image;

[0034] Use feature point detection algorithm to extract feature points in the image;

[0035] Use the local binary pattern method to analyze the texture features of the image;

[0036] Extract shape features of objects in images;

[0037] The extracted feature points, texture features and shape features are converted into numerical vectors as input data;

[0038] Use the pre-trained YOLO model for target detection to obtain the target obstacle type.

[0039] In a second aspect, the present invention provides an obstacle state recognition device in an automatic driving environment of a tourist park, comprising a motion model setting module, a data acquisition module, an obstacle recognition module, a path prediction module and an obstacle avoidance module.

[0040] The motion model setting module is used to set corresponding motion models for different target obstacles;

[0041] The data acquisition module is used to collect environmental data and motion parameters in real time through a multi-sensor module mounted on the vehicle;

[0042] The obstacle identification module is used to pre-process and extract features from the environmental data to identify the target obstacle type;

[0043] The path prediction module is used to obtain the predicted path of each obstacle based on the motion model and motion parameters;

[0044] The obstacle avoidance module is used to output the predicted path to the control system of the vehicle to perform obstacle avoidance operations.

[0045] The present invention provides a method and device for identifying the state of obstacles in an automatic driving environment of a tourist park. First, for different types of target obstacles (such as pedestrians, bicycles, other vehicles, etc.) appearing in the tourist park, corresponding motion models are pre-set. These models are usually based on known behavior patterns, such as pedestrian walking speed range, direction change frequency and other characteristics, as well as the acceleration and deceleration capabilities of vehicles. A variety of sensors are usually installed on the automatic driving vehicle, and these devices can collect data about the vehicle's surrounding environment in real time, including the position, shape, size and dynamic information (such as speed, direction) of the obstacles. Next, the system will pre-process the collected raw data, remove noise or irrelevant information, and extract features that help identify the type of obstacles. For example, the outline of pedestrians is identified from the picture taken by the camera through image processing technology, or the distance and relative speed of the vehicle are calculated using radar echo signals. After obtaining enough information, the algorithm will determine the specific category of the obstacle based on the extracted features, and combine the previously set motion model and the real-time detected motion parameters (such as current position, speed, etc.) to predict the movement trajectory of each obstacle in the future. Finally, the system will send the above prediction results to the control system of the automatic driving vehicle. The control system formulates reasonable driving strategies based on this information, such as adjusting the vehicle speed and steering angle, to avoid collisions with any potential obstacles, ensuring the safety of passengers and reaching their destination smoothly. This method uses advanced sensing technology and artificial intelligence algorithms to ensure the experience of tourists while greatly improving the intelligence level and safety of the transportation system in the tourist park. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 It is a flow chart of a method for identifying obstacle status in an automatic driving environment of a sightseeing park according to the present invention.

[0048] Figure 2 It is a flow chart of setting corresponding motion models for different target obstacles of the present invention.

[0049] Figure 3 It is a flow chart of the present invention for collecting environmental data and motion parameters in real time through a multi-sensor module mounted on a vehicle.

[0050] Figure 4The present invention is a flow chart of using radar to scan the surrounding environment and determine whether an obstacle is within a first preset distance.

[0051] Figure 5 It is a flow chart of the present invention that when an obstacle is within a first preset distance, a steerable camera is started to rotate to a corresponding direction to continuously collect image data at the obstacle.

[0052] Figure 6 It is a flow chart of the present invention for preprocessing and extracting features of the environmental data to identify the type of target obstacle.

[0053] Figure 7 It is a flow chart of the present invention for obtaining the predicted path of each obstacle based on the motion model and motion parameters. DETAILED DESCRIPTION

[0054] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0055] See also Figure 1 to Figure 7 The present invention provides a method for identifying obstacle states in an automatic driving environment of a sightseeing park, comprising:

[0056] S101 sets corresponding motion models for different target obstacles;

[0057] The target obstacles include adults, minors, vehicles and fixed objects.

[0058] The specific steps include:

[0059] S201 obtains historical speed data of the target obstacle;

[0060] This process involves using sensor data, video analysis or other forms of data collection technology to obtain a record of the speed at which an object has moved over a period of time.

[0061] S202 obtains the behavior of the target obstacle based on the historical speed data and performs behavior marking;

[0062] The system will analyze the behavior patterns of the target obstacles based on the historical speed data obtained from S201 and label or classify each pattern. For example, if an adult target is detected moving in a straight line at a constant speed, its behavior will be marked as 'walking'; if a car is detected accelerating suddenly, its behavior will be marked as 'accelerating'.

[0063] S203 trains a support vector machine model based on the behavior markers and historical speed data to obtain a motion model.

[0064] The specific way in which the system uses behavioral markers and historical speed data to train a support vector machine (SVM) model is as follows:

[0065] The system collects data from various sensors, such as radar, LIDAR, cameras, etc., which contain information such as the location and speed of target obstacles. These data are preprocessed, including cleaning (removing noise), standardization (making the data comparable), feature extraction (extracting useful information from the raw data), etc.

[0066] Then, based on the historical speed data and other relevant information (such as acceleration, direction change, etc.), the system constructs a feature vector. The feature vector is part of the input of the machine learning model. It contains key information for distinguishing the behavior patterns of different obstacles. The behavior label is used as the target variable of supervised learning and together with the feature vector, it constitutes the training set.

[0067] Use the labeled training set data to train the SVM model. SVM is a supervised learning model, mainly used for classification and regression analysis. Here, its main task is to learn how to correctly predict the behavior pattern of the target obstacle based on the feature vector.

[0068] SVM tries to find a hyperplane that maximizes the interval between two types of data. For multi-class problems, it can be solved in a one-to-many or many-to-many way.

[0069] During the training process, it is necessary to adjust the parameters of SVM, such as the selection of kernel function and penalty coefficient C, to optimize the model performance.

[0070] Use cross-validation to evaluate the generalization ability of the model to ensure that the model not only performs well on the training data, but also maintains high accuracy on future unseen data.

[0071] After finalizing the model parameters, an independent test set was used to evaluate the model’s performance.

[0072] When the model reaches satisfactory performance indicators, it can be deployed in practical applications to predict the future motion trajectory of target obstacles in real time.

[0073] Through the above steps, the support vector machine model can learn to predict the future motion state of obstacles based on their historical behavior characteristics, thus providing a basis for subsequent decision-making. For example, in an autonomous driving scenario, such predictions can help vehicles plan ahead and avoid collision risks.

[0074] S102 collects environmental data and motion parameters in real time through a multi-sensor module mounted on the vehicle;

[0075] The specific steps include:

[0076] S301 uses radar to scan the surrounding environment and determine whether the obstacle is within a first preset distance;

[0077] The specific steps include:

[0078] S401 activates the radar sensor to start emitting electromagnetic waves and receiving reflected signals;

[0079] In this step, the radar system is first activated. The transmitter inside the radar sensor starts working, generating and emitting electromagnetic waves. These electromagnetic waves propagate through the air and are reflected back when they come into contact with any object (such as an obstacle). The radar receiver is responsible for capturing these reflected signals for further processing.

[0080] S402 evaluates the strength of the reflected signal to determine the distance and angle of the reflection source;

[0081] Once the receiver has captured the reflected signals, the next step is to evaluate the strength of these signals. Signal strength is related to the distance from the obstacle, because the signal gradually attenuates during propagation. In addition, signal strength can also help determine the approximate direction of the reflection source. By comparing the signal strength received in different directions, the relative angle of the reflection source can be estimated.

[0082] S403 calculates the distance of the target by the time delay of the radar echo;

[0083] The radar system calculates the distance of the target by measuring the time delay between the transmitted signal and the received reflected signal. Since the speed of electromagnetic waves is known (close to the speed of light), the time delay is proportional to the distance. Through simple mathematical calculations, the distance between the target and the radar can be obtained.

[0084] The S404 uses the Doppler effect to estimate the angle of a target, i.e. the direction of the target relative to the radar;

[0085] In addition to using time delay to calculate distance, radar also uses the Doppler effect to estimate the relative speed and direction of the target. The Doppler effect refers to the change in the frequency of the wave received by the observer when there is relative motion between the wave source and the observer. Radar determines whether the target is approaching or moving away from the radar, as well as its direction of movement by analyzing the frequency change of the reflected signal.

[0086] S405 calculates the difference between the detected obstacle distance and the first preset distance.

[0087] After completing all the above steps, the system will have an exact measurement of the distance to the obstacle. Next, the system needs to compare this measurement with the preset safety distance. If the detected distance is less than or equal to the preset safety distance, it means that the obstacle has entered an area of ​​concern. The system will calculate the difference between the two to decide whether further action is needed, such as slowing down or changing the route.

[0088] S302: when the obstacle is within the first preset distance, starting the steerable camera to rotate to a corresponding direction to continuously collect image data at the obstacle;

[0089] The specific steps include:

[0090] S501 The control system sends a rotation instruction to the steerable camera, instructing the camera to turn to the direction where the obstacle is located;

[0091] In this step, the vehicle's central control system (part of the autonomous driving system) detects information from the radar and confirms that the obstacle is indeed within the preset safe distance. The control system then sends a clear command to the steerable camera, asking the camera to adjust its orientation to the direction of the obstacle. This command is usually issued by the electronic control unit (ECU), which is transmitted to the camera's drive unit via the vehicle network (such as the CAN bus).

[0092] S502 adjusts the position of the camera according to the angle information provided by the radar so that the camera points to the obstacle;

[0093] When the camera receives a rotation command, it will use the built-in motor or actuator to adjust its position. The camera's position adjustment is based on the angle information provided by the radar. The radar has calculated the position of the obstacle relative to the vehicle in the previous step, including the horizontal and vertical angles. This data is used to guide the rotation of the camera to ensure that its lens can accurately point to the location of the obstacle. This process involves complex algorithms to ensure that the camera can quickly and accurately aim at the target even when the vehicle is moving.

[0094] S503 Aim the camera at the obstacle and start the camera to collect images.

[0095] Once the camera successfully aims at the obstacle, the system will start the camera to start image acquisition. At this time, the camera will continuously take pictures of the obstacle or record video streams for subsequent image processing and analysis. These image data will be transmitted to the central control system for further processing to identify the type, size, shape and other information of the obstacle, and make appropriate responses accordingly, such as adjusting the vehicle's driving path or slowing down.

[0096] Through this series of detailed steps, the system can not only detect surrounding obstacles in a timely manner, but also further enhance its understanding of the environment through visual information, thereby improving the safety and reliability of the autonomous driving system.

[0097] S303 uses a laser rangefinder to measure the motion parameters of obstacles.

[0098] The system first activates the laser rangefinder (commonly known as laser radar or LiDAR) installed on the vehicle. Laser rangefinder is a technology that determines distance by emitting a laser beam and measuring its reflection time. The activation process involves the power management unit supplying power to the laser rangefinder and initializing its control circuits and sensors.

[0099] When a laser rangefinder starts working, it emits short pulses of laser beams. These laser beams propagate through the air and are reflected back when they encounter obstacles. The laser rangefinder is equipped with a special receiver to capture the returning laser signal. The receiver records the time it takes for the signal to travel back and forth and calculates the distance based on the speed of light.

[0100] When the receiver captures the reflected signal, the system calculates the time it takes for the laser pulse to be emitted and received. Since the speed of the laser is known (the speed of light), the distance between the obstacle and the vehicle can be easily calculated by multiplying the time by half the speed of light. This is because the signal goes back and forth once, so we take half the distance.

[0101] Modern laser rangefinders often use multi-beam or multi-band technology, which means that it can emit multiple laser beams at the same time or use lasers of different frequencies to cover a larger range or obtain more detailed data. This not only improves the accuracy of the measurement, but also provides point cloud data in three-dimensional space, allowing the system to build a three-dimensional model of obstacles.

[0102] In order to obtain the motion parameters of the obstacle, the laser rangefinder will continuously scan the surrounding environment. By comparing the measurement results at different time points, the system can analyze the dynamic information of the obstacle, such as speed, direction and even acceleration. For example, if the distance between an obstacle and the vehicle is decreasing after several consecutive measurements, it can be inferred that the obstacle is approaching the vehicle.

[0103] Finally, the system will fuse the data obtained by the laser rangefinder with the data from other sensors (such as radar, camera, etc.). The purpose of this is to combine the advantages of each sensor and eliminate the errors existing in a single sensor, so as to obtain more reliable and comprehensive environmental perception information. These processed data will be used in the decision-making algorithm to help the vehicle make appropriate response measures, such as adjusting the driving speed or changing lanes.

[0104] S103 performs preprocessing and feature extraction on the environmental data to identify the target obstacle type;

[0105] The specific steps include:

[0106] S601 converts the image data into a grayscale image;

[0107] In this step, the system first converts the color image into a grayscale image. This is because in many cases, the grayscale image is sufficient to provide enough information for subsequent processing, and it is faster and more efficient to process. A grayscale image is a single-channel image consisting of the brightness values ​​of pixels, and the value of each pixel represents the brightness level at that location.

[0108] S602 adjusts the grayscale image to a preset size;

[0109] Next, the system will resize the grayscale image to a fixed size. This is done to ensure that all image data is consistent in subsequent processing and is easier for the model to handle. The preset size is usually determined based on the algorithm requirements or hardware limitations of the subsequent use.

[0110] S603 uses a Sobel algorithm to detect edge-protruding object contours in the resized grayscale image;

[0111] In this step, the system applies the Sobel algorithm to detect edges in the image. The Sobel algorithm is a commonly used edge detection technique that enhances the pixel values ​​in the edge area by performing a convolution operation on the image. This can highlight the outline of objects in the image and facilitate subsequent feature extraction.

[0112] S604 extracts feature points in the image using a feature point detection algorithm;

[0113] Feature point detection algorithms (such as SIFT, SURF, or ORB) are used to find unique key points in an image. These feature points are usually located at corners or texture-rich areas in the image. By extracting feature points, the system can better understand the structural information in the image.

[0114] S605 uses a local binary pattern method to analyze texture features of the image;

[0115] Local binary patterns describe texture features by comparing pixel values ​​in local areas of an image to form binary patterns. This method can capture subtle texture changes in an image and is very effective in distinguishing different types of surfaces.

[0116] S606 extracts shape features of objects in the image;

[0117] At this stage, the system further analyzes the shape of the object in the image and extracts shape features. These features include but are not limited to geometric properties such as the object's outline, area, and perimeter. Shape features are very important for identifying different types of target obstacles (such as pedestrians, vehicles, etc.).

[0118] S607 converts the extracted feature points, texture features and shape features into numerical vectors as input data;

[0119] After extracting the above features, the system will integrate them into a numerical vector. This vector contains all the useful feature information and can be used as input data for machine learning models. Numerical vectorization makes it easy for features to be processed and stored by computers.

[0120] S608 uses the pre-trained YOLO model to perform target detection to obtain the target obstacle type.

[0121] Finally, the system uses a pre-trained YOLO (You Only Look Once) model to identify the target obstacles in the image. YOLO is a real-time object detection system that can predict object bounding boxes and their class probabilities directly from the input image. In this way, the system can quickly and accurately identify different types of obstacles in the image and provide necessary information for subsequent decision-making.

[0122] S104 obtains predicted paths of various obstacles based on the motion model and motion parameters.

[0123] The specific steps include:

[0124] S701 converts the predicted paths of all target obstacles into the same reference coordinate system;

[0125] In this step, the system must first ensure that the predicted paths of all target obstacles are unified into the same reference coordinate system. This is because different sensors or data sources have different coordinate system definitions. If they are not converted, the data will be inconsistent and affect the accuracy of subsequent processing. The conversion process involves coordinate transformation formulas, such as translation and rotation, to ensure that all path data are compared in the same reference frame.

[0126] S702 superimposes and displays the predicted paths of all obstacles in the same coordinate system;

[0127] Once all the path data is converted to the same reference coordinate system, the next step is to overlay the path data. The purpose of this is to allow the system to intuitively see the future position and relationship of each obstacle. Overlay display can be achieved by drawing two-dimensional or three-dimensional graphics to make each path clearly visible.

[0128] S703 uses a graphical interface to display the predicted path of each target obstacle;

[0129] In this step, the system uses a graphical user interface (GUI) to display the predicted path for each target obstacle. The GUI can be a dedicated software interface or integrated into the operating system of the autonomous vehicle. Through the graphical interface, not only can the path be displayed, but other relevant information can also be attached, such as the obstacle type, the uncertainty area of ​​the predicted path, etc., so that the operator or system can better understand and use this information.

[0130] S704 plans the driving path of the autonomous driving vehicle according to the predicted paths of each target obstacle.

[0131] This process usually involves advanced path planning algorithms that take into account multiple factors, including but not limited to road conditions, traffic rules, predicted paths of obstacles, and the behavior of other traffic participants. The goal of path planning is to ensure that the vehicle can choose the optimal driving route while ensuring safety and avoid conflicts with any obstacles.

[0132] S105 outputs the predicted path to the vehicle's control system for obstacle avoidance operations.

[0133] S801 packages the predicted path data into a control command format

[0134] At this stage, the system first needs to convert the predicted path data into a format suitable for control system processing. This involves packaging the predicted path information into a specific data packet format, which contains all the necessary information required by the vehicle control system, such as target position, speed, acceleration, etc. The data packaging process also needs to consider the timeliness and accuracy of data transmission to ensure that the path information can be transmitted to the control system in a timely and error-free manner.

[0135] S802 transmits data packets to the control system via the vehicle communication network.

[0136] The packaged data packets need to be transmitted to the vehicle's control system through the vehicle communication network (such as CAN bus, LIN bus or other dedicated vehicle network). This step ensures that the path information can be accurately transmitted from the processing unit (such as computing platform, ECU, etc.) to the controller that performs obstacle avoidance operations. The design of the vehicle communication network needs to ensure the real-time and reliability of data transmission to prevent control errors caused by network delays or data loss.

[0137] S803 control system parses path data packets and executes obstacle avoidance logic

[0138] When the control system receives the path data packets, it will parse them and extract the necessary path information. The software or firmware inside the control system will execute the predetermined obstacle avoidance logic based on the parsed information. The obstacle avoidance logic includes but is not limited to:

[0139] Replan the vehicle's path based on the predicted path to avoid obstacles;

[0140] Adjust the vehicle's speed to ensure safe movement around obstacles;

[0141] Change the direction of the vehicle by adjusting the angle of the steering wheel;

[0142] The brake system is activated when necessary to avoid a collision.

[0143] S804 performs obstacle avoidance action

[0144] After completing the analysis and logical processing of the path information, the control system will send instructions to the vehicle's actuators (such as motors, hydraulic systems, brake systems, etc.) to perform specific obstacle avoidance actions. These actions include deceleration, lane change, parking, etc., depending on the requirements of the predicted path and the current driving environment.

[0145] S805 monitors obstacle avoidance effect and provides feedback

[0146] During the entire obstacle avoidance operation, the control system will continue to monitor the obstacle avoidance effect and adjust the control strategy according to the actual situation. If the obstacle avoidance operation is successful, the control system will continue to perform the original planned driving task; if it encounters new obstacles or unexpected situations, the system needs to re-evaluate the environment and output new predicted path information again to enter the next cycle.

[0147] Through the above detailed steps, the system can not only effectively transmit the predicted path information to the vehicle's control system, but also ensure that the vehicle can perform obstacle avoidance operations in a timely and accurate manner based on this information, thereby ensuring driving safety.

[0148] Second embodiment

[0149] The present invention provides an obstacle state recognition device in an automatic driving environment of a tourist park, comprising a motion model setting module, a data acquisition module, an obstacle recognition module, a path prediction module and an obstacle avoidance module, wherein the motion model setting module is used to set corresponding motion models for different target obstacles; the data acquisition module is used to collect environmental data and motion parameters in real time through a multi-sensor module mounted on a vehicle; the obstacle recognition module is used to preprocess and extract features of the environmental data to identify the type of target obstacle; the path prediction module is used to obtain a predicted path of each obstacle based on the motion model and motion parameters; and the obstacle avoidance module is used to output the predicted path to the control system of the vehicle to perform an obstacle avoidance operation.

[0150] In this embodiment, the motion model setting module establishes corresponding motion models for different types of obstacles (such as adults, minors, vehicles and fixed objects). These models are trained by machine learning algorithms (such as support vector machines SVM) based on the behavior patterns and historical speed data of obstacles to predict the future motion trajectory of obstacles. This module extracts the behavioral features of obstacles by analyzing historical data and uses these features to train prediction models. In this way, the system can model the behavior of different types of obstacles and predict their future motion states based on the models.

[0151] The data acquisition module uses a combination of multiple sensors (such as radar, lidar, camera, etc.) installed on the autonomous vehicle to collect data about the surrounding environment and the motion parameters of obstacles in real time. These sensors can provide information about the location, speed, direction, etc. of obstacles. The radar scans the surrounding environment to determine whether the obstacle is within the preset safe distance; the camera is used to obtain the image data of the obstacle; and the laser rangefinder is used to measure the distance and motion parameters of the obstacle. These data provide basic information for subsequent obstacle identification and path prediction.

[0152] The obstacle recognition module preprocesses and extracts features from the collected environmental data to identify the type of target obstacles. This module enhances image features through a series of image processing techniques (such as grayscale conversion, resizing, edge detection, feature point extraction, etc.), and uses advanced image recognition algorithms (such as YOLO) to classify obstacles. This converts image data into a form that can be analyzed, extracts key features such as edges, textures, and shapes, and then identifies the specific type of obstacles.

[0153] The path prediction module predicts the future path of each obstacle based on the motion model and motion parameters. This module unifies the predicted paths of obstacles into the same reference coordinate system and displays them in a superimposed manner in the same coordinate system to facilitate the overall evaluation of the dynamics of the obstacles. This module uses the previously established motion model and the motion parameters collected in real time to predict the future position and motion trajectory of obstacles, providing a basis for obstacle avoidance operations.

[0154] The obstacle avoidance module transmits the predicted path information to the vehicle's control system for obstacle avoidance operations. The module packages the predicted path data into a control command format and transmits it to the control system through the vehicle communication network. Based on the predicted path data, the module sends instructions to the vehicle's actuators to perform specific obstacle avoidance actions, such as slowing down, changing lanes, or stopping, to ensure safe driving of the vehicle.

[0155] Through the collaboration of the above five modules, the entire device realizes real-time recognition and effective obstacle avoidance of obstacles in the autonomous driving environment in the tourist park. From data collection, obstacle recognition, path prediction to the final obstacle avoidance operation, each module works closely together to ensure that the autonomous driving vehicle can safely and efficiently cope with various complex road conditions, improving traffic safety and passenger experience in the tourist park.

[0156] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for identifying obstacle states in an autonomous driving environment in a tourist park. It is characterized in that Including: setting corresponding motion models for different target obstacles; Collect environmental data and motion parameters in real time through multi-sensor modules mounted on vehicles; Preprocessing and feature extracting the environmental data to identify the target obstacle type; Get the predicted path of each obstacle based on the motion model and motion parameters; The predicted path is output to the vehicle's control system for obstacle avoidance.

2. The method for identifying obstacle status in an automatic driving environment of a tourist park according to claim 1, characterized in that: The target obstacles include people, vehicles and fixed objects.

3. The method for identifying obstacle status in an automatic driving environment of a tourist park as claimed in claim 2, characterized in that: The specific steps of setting corresponding motion models for different target obstacles include: Obtain historical speed data of target obstacles; Obtain the behavior of the target obstacle based on historical speed data and mark the behavior; The motion model is obtained by training the support vector machine model based on the behavior markers and historical speed data.

4. The method for identifying obstacle status in an automatic driving environment of a tourist park as claimed in claim 3, characterized in that: The specific steps of collecting environmental data and motion parameters in real time through the multi-sensor module mounted on the vehicle include: Use radar to scan the surrounding environment and determine whether the obstacle is within the first preset distance; When the obstacle is within the first preset distance, the steerable camera is started to rotate to the corresponding direction to continuously collect image data at the obstacle; A laser rangefinder is used to measure the motion parameters of obstacles.

5. The method for identifying obstacle status in an automatic driving environment of a tourist park as claimed in claim 4, characterized in that: The specific steps of using radar to scan the surrounding environment and determine whether an obstacle is within the first preset distance include: Activate the radar sensor so that it starts to emit electromagnetic waves and receive the reflected signals; Evaluate the strength of the reflected signal to determine the distance and angle of the reflection source; Calculate the distance to the target by the time delay of the radar echo; Use the Doppler effect to estimate the target's angle, i.e. the direction of the target relative to the radar; The difference between the detected obstacle distance and the first preset distance is calculated.

6. The method for identifying obstacle status in an automatic driving environment of a tourist park as claimed in claim 5, characterized in that: The specific steps of starting the steerable camera to rotate to a corresponding direction to continuously collect image data at the obstacle when the obstacle is within the first preset distance include: The control system sends a rotation instruction to the steerable camera, instructing the camera to turn to the direction where the obstacle is located; Adjust the position of the camera according to the angle information provided by the radar so that the camera points to the obstacle; The camera is pointed at the obstacle and starts to collect images.

7. The method for identifying obstacle status in an automatic driving environment of a tourist park as claimed in claim 6, characterized in that: The specific steps of preprocessing and feature extracting the environmental data to identify the target obstacle type include: Convert image data to grayscale image; Resize grayscale images to a preset size; Use the Sobel algorithm to detect edge-prominent object contours in the resized grayscale image; Use feature point detection algorithm to extract feature points in the image; Use the local binary pattern method to analyze the texture features of the image; Extract shape features of objects in images; The extracted feature points, texture features and shape features are converted into numerical vectors as input data; Use the pre-trained YOLO model for target detection to obtain the target obstacle type.

8. An obstacle state recognition device in an automatic driving environment of a sightseeing park, applied to an obstacle state recognition method in an automatic driving environment of a sightseeing park as claimed in any one of claims 1 to 7, characterized in that: It includes motion model setting module, data acquisition module, obstacle recognition module, path prediction module and obstacle avoidance module. The motion model setting module is used to set corresponding motion models for different target obstacles; The data acquisition module is used to collect environmental data and motion parameters in real time through a multi-sensor module mounted on the vehicle; The obstacle identification module is used to pre-process and extract features from the environmental data to identify the target obstacle type; The path prediction module is used to obtain the predicted path of each obstacle based on the motion model and motion parameters; The obstacle avoidance module is used to output the predicted path to the control system of the vehicle to perform obstacle avoidance operations.

Images