Obstacle avoidance scene detection method, device, equipment and storage medium

Abstract

The present invention relates to the field of automatic driving detection technology, and in particular to a detection method, device, equipment and storage medium for obstacle avoidance scenes. By acquiring the driving data of the main vehicle during driving, the target obstacles in the surrounding environment and the position information of the target obstacles, the predicted driving trajectory of the main vehicle in the next time period in the current driving direction is calculated based on the driving data, and the first distance of the predicted driving trajectory relative to the target obstacle and the second distance of the main vehicle relative to the target obstacle are calculated based on the driving data, the position information of the target obstacle and the predicted driving trajectory. Based on the first distance and the second distance, it is determined whether the scene corresponding to the road test data is an obstacle avoidance scene. The problem that obstacles cannot be effectively screened accurately and the accuracy of obstacle avoidance scene detection is low in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving detection technology, and in particular to a detection method, device, equipment and storage medium for obstacle avoidance scenarios. Background Art

[0002] With the development of autonomous driving technology, how to effectively detect and judge relevant driving scenarios and accident scenarios is an important issue in this technical field during the process of autonomous driving and analysis and feedback of autonomous driving case data.

[0003] Currently, in follow-up feedback from autonomous driving road tests, annotators or engineers will manually label the types of problems that occurred during the road tests, such as the main vehicle being too close to an obstacle. Annotators or engineers will perform annotations based on film and television footage, driver recordings, and simulations of the vehicle's internal and external conditions. However, this manual labeling and judgment method is difficult to achieve standardization and accuracy requirements in high-speed scenarios, especially when the current distance between the main vehicle and the obstacle is large, but there is still a risk of collision. The judgment criteria for such scenarios vary, and most rely on feeling. The above reasons lead to the problem that the existing technology is unable to effectively screen obstacles and is unable to effectively detect scenes where the distance to obstacles is too close.

[0004] Therefore, there is an urgent need to provide a detection solution that can effectively screen obstacles and effectively detect scenes that are too close to obstacles. Summary of the Invention

[0005] The main purpose of this invention is to solve the problem of low recognition efficiency and low accuracy in autonomous driving tests due to inconsistent obstacle avoidance scene recognition standards.

[0006] A first aspect of the present invention provides a method for detecting an obstacle avoidance scenario, which includes: obtaining road test data of a main vehicle during driving, wherein the road test data includes driving data, target obstacles in the surrounding environment, and position information of the target obstacles; calculating a predicted driving trajectory of the main vehicle in the next time period in the current driving direction based on the driving data; calculating a first distance of the predicted driving trajectory relative to the target obstacle based on the position information of the target obstacle and the predicted driving trajectory; calculating a second distance of the main vehicle relative to the target obstacle based on the position information of the target obstacle and the driving data; judging whether the first distance and the second distance meet preset obstacle avoidance scenario judgment conditions, and determining whether the scene corresponding to the road test data is an obstacle avoidance scenario based on the judgment result.

[0007] Optionally, in a first implementation manner of the first aspect of the present invention, obtaining the road test data of the main vehicle during driving includes: obtaining a semantic map and driving data corresponding to the main vehicle, wherein the driving data includes the current position information, driving speed and driving direction of the main vehicle; performing semantic parsing on the semantic map to identify all obstacles therein; calculating a third distance between each obstacle and both sides of the road, and based on the third distance, filtering out obstacles that are not within the curb range of the road from all the obstacles to obtain a target obstacle; calculating the coordinates of the target obstacle in the semantic map to obtain the position information of the target obstacle.

[0008] Optionally, in a second implementation manner of the first aspect of the present invention, calculating the first distance of the predicted driving trajectory relative to the target obstacle based on the position information of the target obstacle and the predicted driving trajectory includes: calculating the point position information of the main vehicle at the end point of the next time period based on the driving speed of the main vehicle and the predicted driving trajectory; and calculating the longitudinal distance and lateral distance of the main vehicle relative to the target obstacle at the end point of the next time period based on the position information of the target obstacle and the point position information.

[0009] Optionally, in a third implementation manner of the first aspect of the present invention, calculating the second distance of the main vehicle relative to the target obstacle based on the position information of the target obstacle and the driving data includes: calculating the relative distance between the main vehicle and the target obstacle based on the current position information and the position information of the target obstacle; identifying whether the target obstacle is stationary; if the target obstacle is stationary, calculating a first displacement of the main vehicle in the next unit time based on the driving speed; calculating the second distance of the main vehicle relative to the target obstacle based on the first displacement and the relative distance; if the target obstacle is not stationary, obtaining the moving speed of the target obstacle, and calculating the second displacement of the main vehicle in the next unit time based on the driving speed and the moving speed; and calculating the second distance of the main vehicle relative to the target obstacle based on the second displacement and the relative distance.

[0010] Optionally, in a fourth implementation manner of the first aspect of the present invention, the judgment of whether the first distance and the second distance meet the preset obstacle avoidance scene judgment condition is obtained, and a judgment result is obtained, and based on the judgment result, whether the scene corresponding to the road test data is an obstacle avoidance scene is determined, including: judging whether the longitudinal distance and the lateral distance meet the trajectory obstacle avoidance condition; judging whether the second distance meets the position obstacle avoidance condition; if at least one of the longitudinal distance, the lateral distance and the second distance meets the condition, then the scene corresponding to the road test data is determined to be an obstacle avoidance scene; if the longitudinal distance, the lateral distance and the second distance are not met, then the scene corresponding to the road test data is determined to be a non-obstacle avoidance scene.

[0011] Optionally, in the fifth implementation manner of the first aspect of the present invention, after determining that at least one of the longitudinal distance, the lateral distance and the second distance is satisfied, it also includes: detecting whether the target obstacle in the next frame of road test data of the main vehicle meets the obstacle avoidance scene judgment condition; if so, determining that the scene corresponding to the road test data is an obstacle avoidance scene; if not, determining that the scene corresponding to the road test data is a non-obstacle avoidance scene.

[0012] Optionally, in a sixth implementation of the first aspect of the present invention, the method of calculating the predicted driving trajectory of the main vehicle in the next time period in the current driving direction based on the driving data includes: inputting the driving data into a pre-trained neural network model, wherein the neural network model includes an input layer, an embedding layer, a hidden layer and an output layer; converting the driving data into a trajectory vector matrix through the input layer, and outputting it to the embedding layer; converting the trajectory vector matrix into a corresponding embedding expression matrix through the embedding layer, and outputting it to the hidden layer; calculating the label vector corresponding to the embedding expression matrix through the hidden layer, and outputting it to the output layer; calling a classification algorithm through the output layer to calculate the classification probability of the embedding expression matrix and its corresponding label vector, and outputting a prediction result; and embedding the prediction result into the semantic map in a graphical manner to obtain the predicted driving trajectory of the main vehicle in the future time period.

[0013] A second aspect of the present invention provides a detection device for an obstacle avoidance scene, comprising: an acquisition module for acquiring road test data of a main vehicle during driving, wherein the road test data includes driving data, target obstacles in the surrounding environment, and position information of the target obstacles; a prediction module for calculating, based on the driving data, a predicted driving trajectory of the main vehicle in the next time period in the current driving direction; a first calculation module for calculating a first distance of the predicted driving trajectory relative to the target obstacle based on the position information of the target obstacle and the predicted driving trajectory; a second calculation module for calculating a second distance of the main vehicle relative to the target obstacle based on the position information of the target obstacle and the driving data; a judgment module for judging whether the first distance and the second distance meet a preset obstacle avoidance scene judgment condition, and determining whether the scene corresponding to the road test data is an obstacle avoidance scene based on the judgment result.

[0014] Optionally, in a first implementation manner of the second aspect of the present invention, the acquisition module includes: a data acquisition unit, used to acquire the semantic map and driving data corresponding to the main vehicle, wherein the driving data includes the current position information, driving speed and driving direction of the main vehicle; a semantic parsing unit, used to perform semantic parsing on the semantic map and identify all obstacles therein; a first calculation unit, used to calculate a third distance between each obstacle and both sides of the road, and based on the third distance, filter out obstacles that are not within the curb range of the road from all the obstacles to obtain a target obstacle; and a second calculation unit, used to calculate the coordinates of the target obstacle in the semantic map to obtain the position information of the target obstacle.

[0015] Optionally, in a second implementation of the second aspect of the present invention, the first calculation module includes: a third calculation unit, used to calculate the point position information of the main vehicle at the end point of the next time period based on the driving speed of the main vehicle and the predicted driving trajectory; a fourth calculation unit, used to calculate the longitudinal distance and lateral distance of the main vehicle relative to the target obstacle at the end point of the next time period based on the position information of the target obstacle and the point position information.

[0016] Optionally, in a third implementation manner of the second aspect of the present invention, the second calculation module includes: a fifth calculation unit, configured to calculate the relative distance between the main vehicle and the target obstacle based on the current position information and the position information of the target obstacle; a stationary judgment unit, configured to identify whether the target obstacle is in a stationary state; a sixth calculation unit, configured to calculate a first displacement of the main vehicle in the next unit time based on the driving speed when the target obstacle is in a stationary state; and calculate a second distance of the main vehicle relative to the target obstacle based on the first displacement and the relative distance; a seventh calculation unit, configured to obtain the moving speed of the target obstacle if the target obstacle is not in a stationary state, and calculate the second displacement of the main vehicle in the next unit time based on the driving speed and the moving speed; and an eighth calculation unit, configured to calculate the second distance of the main vehicle relative to the target obstacle based on the second displacement and the relative distance.

[0017] Optionally, in a fourth implementation manner of the second aspect of the present invention, the judgment module includes: a first judgment unit, used to judge whether the longitudinal distance and the lateral distance meet the trajectory obstacle avoidance condition; a second judgment unit, used to judge whether the second distance meets the position obstacle avoidance condition; a first determination unit, used to determine that the scene corresponding to the road test data is an obstacle avoidance scene when at least one of the longitudinal distance, the lateral distance and the second distance meets the condition; and a second determination unit, used to determine that the scene corresponding to the road test data is a non-obstacle avoidance scene when none of the longitudinal distance, the lateral distance and the second distance meets the condition.

[0018] Optionally, in a fifth implementation of the second aspect of the present invention, the obstacle avoidance scene detection device also includes a second judgment module, which is specifically used to detect whether the target obstacle in the next frame of road test data of the main vehicle meets the obstacle avoidance scene judgment condition after judging that at least one of the longitudinal distance, the lateral distance and the second distance is satisfied; if so, it is determined that the scene corresponding to the road test data is an obstacle avoidance scene; if not, it is determined that the scene corresponding to the road test data is a non-obstacle avoidance scene.

[0019] Optionally, in a sixth implementation of the second aspect of the present invention, the prediction module includes: a data input unit, used to input the driving data into a pre-trained neural network model, wherein the neural network model includes an input layer, an embedding layer, a hidden layer and an output layer; an input layer unit, used to convert the driving data into a trajectory vector matrix through the input layer, and output it to the embedding layer; an embedding layer unit, used to convert the trajectory vector matrix into a corresponding embedding expression matrix through the embedding layer, and output it to the hidden layer; a hidden layer unit, used to calculate the label vector corresponding to the embedding expression matrix through the hidden layer, and output it to the output layer; an output layer unit, used to call a classification algorithm through the output layer to perform classification probability calculation on the embedding expression matrix and its corresponding label vector, and output a prediction result; a trajectory generation unit, used to embed the prediction result into the semantic map in a graphical manner to obtain the predicted driving trajectory of the main vehicle in a certain period of time in the future.

[0020] A third aspect of the present invention provides a computer device comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes each step of the above-mentioned obstacle avoidance scene detection method.

[0021] A fourth aspect of the present invention provides a computer-readable storage medium, which stores instructions that, when executed on a computer, enable the computer to execute the various steps of the above-mentioned obstacle avoidance scene detection method.

[0022] In the technical solution of the present invention, the method specifically obtains driving data of the main vehicle during driving, target obstacles in the surrounding environment, and position information of the target obstacles, calculates a predicted driving trajectory of the main vehicle in the next time period in the current driving direction based on the driving data, calculates a first distance of the predicted driving trajectory relative to the target obstacle and a second distance of the main vehicle relative to the target obstacle based on the driving data, and determines whether the scene corresponding to the road test data is an obstacle avoidance scene based on the first distance and the second distance; thereby solving the problems in the prior art of being unable to accurately screen obstacles and having a low accuracy rate in obstacle avoidance scene detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 1 is a schematic diagram of a first embodiment of a detection method for an obstacle avoidance scene according to an embodiment of the present invention;

[0024] Figure 2 2 is a schematic diagram of a second embodiment of a detection method for an obstacle avoidance scene in an embodiment of the present invention;

[0025] Figure 3 2 is a schematic diagram of a third embodiment of a detection method for an obstacle avoidance scene in an embodiment of the present invention;

[0026] Figure 4 Schematic diagram of an embodiment of a detection device for an obstacle avoidance scene in an embodiment of the present invention;

[0027] Figure 5 Schematic diagram of another embodiment of a detection device for an obstacle avoidance scene according to an embodiment of the present invention;

[0028] Figure 6 FIG. 1 is a schematic diagram of an embodiment of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to solve the problems in the prior art that obstacles cannot be effectively screened and scenes that are too close to obstacles cannot be effectively detected. The present application provides a method, device, equipment and storage medium for detecting obstacle avoidance scenes. The method obtains road test data of the main vehicle during driving, wherein the road test data includes driving data, target obstacles in the surrounding environment and position information of the target obstacles; based on the driving data, the predicted driving trajectory of the main vehicle in a certain time period in the future is calculated; according to the driving data, the position information of the target obstacle and the predicted driving trajectory, it is determined whether the preset too close scene judgment conditions are met; if so, it is determined that the main vehicle is too close to the target obstacle, and the road test data is too close scene data; as described above, by analyzing the road test data and judging the analysis results based on distance, trajectory and speed, it is possible to effectively screen target obstacles and detect scenes that are too close to obstacles; thereby solving the problems in the prior art that obstacles cannot be effectively screened and scenes that are too close to obstacles cannot be effectively detected.

[0030] The terms "first," "second," "third," "fourth," and the like (if any) in the description and claims of the present invention and in the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatus.

[0031] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1In the first embodiment of the obstacle avoidance scene detection method according to the present invention, the steps for implementing the method are as follows:

[0032] 101. Obtaining road test data of the main vehicle during driving;

[0033] In this step, the drive test data includes driving data, target obstacles in the surrounding environment, and location information of the target obstacles;

[0034] This step can be implemented in the following ways:

[0035] Obtaining a semantic map and driving data corresponding to the host vehicle, wherein the driving data includes location information, driving speed, and driving direction corresponding to the host vehicle;

[0036] Parsing the semantic map to obtain position points corresponding to edges on both sides of the road, wherein the position points corresponding to the edges on both sides of the road include at least position points corresponding to the edge on the first side of the road and position points corresponding to the edge on the second side of the road;

[0037] Parsing the position information corresponding to the host vehicle and the position information of the target obstacle to obtain all position points corresponding to the host vehicle and all position points corresponding to the target obstacle;

[0038] Calculate the distance between each position point corresponding to the target obstacle and all position points corresponding to the host vehicle, and select the closest position point corresponding to the target obstacle according to the principle of minimum distance. The minimum distance between the closest position point and all position points corresponding to the host vehicle is recorded as the relative closest distance;

[0039] Calculating the distance between the closest point and the point corresponding to the first side edge of the road, and selecting the smallest distance value therebetween to obtain the minimum distance along the first side edge;

[0040] Calculating the distance between the closest point and the point corresponding to the second side edge of the road, and selecting the smallest distance value therebetween to obtain the minimum distance along the second side edge;

[0041] According to a preset determination method based on the closest distance point, the relative closest distance is compared with the first side edge minimum distance and the second side edge minimum distance respectively to determine whether the target obstacle is located within the two side edges of the road; if so, the target obstacle information is retained; if not, the target obstacle information is discarded;

[0042] A target obstacle in a surrounding environment and location information of the target obstacle are acquired, and drive test data is constructed based on the semantic map, the driving data, the target obstacle in the surrounding environment and the location information of the target obstacle.

[0043] 102. Calculate a predicted driving trajectory of the host vehicle in the next time period in the current driving direction based on the driving data;

[0044] This step can be implemented in the following ways:

[0045] Inputting the driving data into a pre-trained neural network model, wherein the Traj-MTL network model includes an input layer, an embedding layer, a hidden layer, and an output layer;

[0046] Wherein, the neural network model is a Traj-MTL network model;

[0047] Converting the driving data into a trajectory vector matrix through the input layer and outputting it to the embedding layer;

[0048] Converting the trajectory vector matrix into a corresponding embedding expression matrix through the embedding layer and outputting it to the hidden layer;

[0049] Calculate the label vector corresponding to the embedded expression matrix through the hidden layer and output it to the output layer;

[0050] Calling a classification algorithm through the output layer to calculate the classification probability of the embedded expression matrix and its corresponding label vector, and outputting a prediction result; wherein the classification algorithm is a Softmax algorithm;

[0051] The prediction result is embedded into the semantic map in a graphical manner to obtain the predicted driving trajectory of the main vehicle in the next time period in the current driving direction.

[0052] In practical applications, the process of calculating the predicted driving trajectory of the host vehicle in the next time period in the current driving direction based on the driving data can also be achieved by:

[0053] Obtaining each frame of driving data of the main vehicle within a certain period of time in the past;

[0054] Analyzing each frame of driving data and constructing a driving trajectory function based on the relationship between the main vehicle position and time;

[0055] The predicted trajectory fitting value of the host vehicle in the next time period in the current driving direction is calculated by using the driving trajectory function to obtain the predicted driving trajectory of the host vehicle in the next time period in the current driving direction.

[0056] 103. Calculate a first distance of the predicted driving trajectory relative to the target obstacle based on the position information of the target obstacle and the predicted driving trajectory;

[0057] This step can be implemented in the following ways:

[0058] Analyzing the predicted driving trajectory to obtain all position points on the predicted driving trajectory of the host vehicle within the next time period in the current driving direction;

[0059] Analyzing the position information of the target obstacle to obtain all position points corresponding to the target obstacle at a preset time point;

[0060] The distances between all position points on the predicted driving trajectory of the host vehicle in the next time period in the current driving direction and all position points corresponding to the target obstacles are calculated, and the smallest distance value is selected to obtain the first distance.

[0061] 104. Calculate a second distance between the host vehicle and the target obstacle based on the position information of the target obstacle and the driving data;

[0062] This step can be implemented in the following ways:

[0063] The position information and driving data of the target obstacle are parsed to obtain a relative distance vector and a relative speed vector between the host vehicle and the target obstacle. For example, the position coordinates of the host vehicle and the target obstacle are obtained by parsing the position information and driving data of the target obstacle, and a relative distance vector pointing from the position coordinates of the host vehicle to the position coordinates of the target obstacle is constructed. The average speed vector of the host vehicle and the average speed vector of the target obstacle in the previous unit time are obtained by parsing the position information and driving data of the target obstacle, and a relative speed vector is obtained based on vector operations.

[0064] The relative distance of the next time unit is calculated based on the relative distance vector and the relative speed vector to obtain the second distance. For example, the relative displacement vector of the next time unit is calculated based on the relative speed vector, and vector addition operation is performed on the relative distance vector and the relative displacement vector to obtain the relative distance vector corresponding to the next time unit. The modulus of the relative distance vector corresponding to the next time unit is taken to obtain the relative distance of the next time unit, which is the second distance.

[0065] 105. Determine whether the first distance and the second distance meet a preset obstacle avoidance scenario determination condition, and determine whether the scenario corresponding to the drive test data is an obstacle avoidance scenario based on the determination result;

[0066] In this step, the preset obstacle avoidance scenario judgment conditions include a trajectory obstacle avoidance condition and a position obstacle avoidance condition, wherein the trajectory obstacle avoidance condition at least includes a trajectory obstacle avoidance threshold, and the position obstacle avoidance condition at least includes a position obstacle avoidance threshold;

[0067] This step can be implemented in the following ways:

[0068] Determining whether the first distance satisfies the trajectory obstacle avoidance condition, for example, determining whether the first distance is less than the trajectory obstacle avoidance threshold; if so, the condition is satisfied; otherwise, the condition is not satisfied;

[0069] Determine whether the second distance satisfies the position obstacle avoidance condition, for example, determine whether the second distance is less than the position obstacle avoidance threshold, if so, the condition is satisfied; otherwise, the condition is not satisfied;

[0070] If both the trajectory obstacle avoidance condition and the position obstacle avoidance condition are not satisfied, it is determined that the scene corresponding to the drive test data is not an obstacle avoidance scene; otherwise, it is determined that the scene corresponding to the drive test data is an obstacle avoidance scene.

[0071] By implementing the above method, driving data of the main vehicle during driving, target obstacles in the surrounding environment, and position information of the target obstacles are obtained, and a predicted driving trajectory of the main vehicle in the next time period in the current driving direction is calculated based on the driving data. A first distance of the predicted driving trajectory relative to the target obstacle and a second distance of the main vehicle relative to the target obstacle are calculated based on the driving data, the position information of the target obstacle, and the predicted driving trajectory. Based on the first distance and the second distance, it is determined whether the scene corresponding to the drive test data is an obstacle avoidance scene. In the above, by analyzing the position information of the target obstacle, the target obstacle can be effectively screened, and the obstacle avoidance scene can be detected and judged based on the distance calculation of the predicted driving trajectory and the position, thereby solving the problems of the existing technology of the inability to effectively screen obstacles and the low accuracy of obstacle avoidance scene detection.

[0072] See also Figure 2 The second embodiment of the obstacle avoidance scene detection method in the embodiment of the present invention is implemented as follows:

[0073] 201. Obtain vehicle operation data and collected environmental data of the main vehicle during driving;

[0074] In this step, the vehicle operation data includes at least wheel speed, steering parameters and engine operation status, and the collected environmental data includes at least satellite positioning information and visible light images around the main vehicle;

[0075] This step can be implemented in the following ways:

[0076] Acquiring vehicle operation data collected by preset sensors connected to the vehicle power and steering systems, wherein the sensors include at least a motion sensor and a temperature sensor;

[0077] Environmental data collected by a preset environmental information collection device is obtained, wherein the environmental information collection device includes at least a camera, a radar, and a laser ranging sensor.

[0078] 202. Analyze the vehicle operation data and the collected environmental data to obtain road test information;

[0079] In this step, the road test information includes at least the host vehicle position information, the host vehicle speed information, the position information of all obstacles within a preset range, the speed information of all obstacles within the preset range, and the road information.

[0080] 203. Screen all obstacles within a preset range in the drive test information to obtain a target obstacle;

[0081] This step can be implemented in the following ways:

[0082] Analyzing the drive test information to obtain the main vehicle position information, the position information of all obstacles, and the road edge information;

[0083] Extracting characteristic position points on the road edge line according to the road edge information, for example, extracting equidistant position points on the road edge line, wherein the distance between two adjacent position points is 0.1 meters;

[0084] Extracting the closest position of each obstacle to the main vehicle based on the main vehicle position information and the position information of all obstacles;

[0085] Based on the positional relationship between the position point of each obstacle closest to the host vehicle and the characteristic position point on the road edge line, determining whether the position point of each obstacle closest to the host vehicle falls between the road edge lines, for example, constructing a road fitting plane using the equidistant position points falling on the road edge lines, and calculating whether the coordinates of the position point of each obstacle closest to the host vehicle are contained in the road fitting plane; if so, the position point of the obstacle closest to the host vehicle falls between the road edge lines; if not, the position point of the obstacle closest to the host vehicle does not fall between the road edge lines;

[0086] If the obstacle's closest position to the main vehicle falls between the road edge lines, the corresponding obstacle is determined to be the target obstacle;

[0087] If the obstacle's closest position to the host vehicle does not fall between the road edge lines, the corresponding obstacle is determined not to be a target obstacle.

[0088] 204. Determine the value of the distance risk coefficient corresponding to the host vehicle and the target obstacle based on the road test information and output it;

[0089] In this step, the distance risk coefficient includes at least two values: 0 and 1, wherein 0 indicates that the risk of the distance between the host vehicle and the target obstacle being too close is low, and 1 indicates that the risk of the distance between the host vehicle and the target obstacle being too close is high;

[0090] This step can be implemented in the following ways:

[0091] parsing the drive test data to obtain position information of the target obstacle corresponding to two moments separated by a preset time unit, and determining whether the position information of the target obstacle corresponding to the two moments is the same;

[0092] If they are the same, the distances between each position point corresponding to the host vehicle and each position point corresponding to the target obstacle are calculated, the smallest distance value is selected, and a determination is made as to whether it is less than a preset static distance too close threshold. If so, a distance risk coefficient of 1 is output; otherwise, a distance risk coefficient of 0 is output;

[0093] If they are not the same, the position information of the host vehicle at the corresponding moment is obtained, and the position information of the host vehicle and the position information of the target obstacle are analyzed to obtain the relative position of the host vehicle and the target obstacle;

[0094] Based on the relative position, determine whether the target obstacle is located in front of or behind the main vehicle. If the target obstacle is not located in front of or behind the main vehicle, determine whether the distance between the position point corresponding to the main vehicle and the position point corresponding to the target obstacle is less than a preset side distance too close threshold. If so, output a distance risk coefficient of 1; otherwise, output a distance risk coefficient of 0;

[0095] If the target obstacle is located in front of or behind the main vehicle, it is determined whether the distance between the position point corresponding to the main vehicle and the position point corresponding to the target obstacle is less than the preset front-to-back distance too close threshold. If so, the output distance risk coefficient is 1; if not, the output distance risk coefficient is 0.

[0096] 205. Calculate the predicted trajectory of the host vehicle based on the road test information;

[0097] This step can be implemented in the following ways:

[0098] Parsing the drive test data to extract the vehicle position information, vehicle speed information, and road information corresponding to the most recent forward time point and the current time point, respectively, wherein the most recent forward time point is a time point that is earlier than the current time point by a minimum preset time unit, and the vehicle speed information includes speed information and acceleration information;

[0099] Calculating an average velocity vector, an average acceleration vector, and an average position coordinate corresponding to a recent forward time period based on the host vehicle position information, host vehicle speed information, and road information corresponding to the recent forward time point and the current time point, wherein the recent forward time period refers to the time period from the recent forward time point to the current time point;

[0100] The predicted trajectory of the host vehicle is calculated based on the average velocity vector, average acceleration vector and average position coordinate corresponding to the most recent time period.

[0101] 206. Determine and output the value of the predicted trajectory distance risk coefficient corresponding to the host vehicle and the target obstacle based on the predicted trajectory of the host vehicle and the location information of the target obstacle;

[0102] In this step, the predicted trajectory distance risk coefficient includes at least two values: 0 and 1, where 0 indicates that the risk of the distance between the predicted trajectory of the host vehicle and the target obstacle being too close is low, and 1 indicates that the risk of the distance between the predicted trajectory of the host vehicle and the target obstacle being too close is high;

[0103] This step can be implemented in the following ways:

[0104] The predicted trajectory of the host vehicle and the position information of the target obstacle are analyzed, and the time when the predicted trajectory of the host vehicle overlaps with the target obstacle is calculated to obtain the ideal collision time. For example, the position coordinates of all the host vehicle positions and the position coordinates of the target obstacle on the predicted trajectory are obtained, and the time corresponding to the time when all the position coordinates of the host vehicle and the position coordinates of the target obstacle are within one meter of each other is calculated to obtain the ideal collision time;

[0105] Determine whether the ideal collision time is less than a preset time threshold. If so, output a predicted trajectory distance risk coefficient of 1; if not, output a predicted trajectory distance risk coefficient of 0.

[0106] 207. Determine the collision risk judgment result based on the distance risk coefficient and the predicted trajectory distance risk coefficient;

[0107] In this step, the collision risk judgment result includes at least two judgment results: a low risk of collision and a high risk of collision;

[0108] This step can be implemented in the following ways:

[0109] Determining whether the distance risk coefficient is 1, and if so, determining that the collision risk determination result is a high risk of collision;

[0110] If not, determining whether the predicted trajectory distance risk coefficient is 1, and if so, determining that the collision risk judgment result is a high risk of collision;

[0111] If not, the collision risk determination result is determined to be a low risk of collision.

[0112] 208. Output collision risk judgment result;

[0113] This step can be implemented in the following ways:

[0114] Outputting the collision risk assessment result through the UI interface. For example, when the collision risk assessment result indicates a high collision risk, a collision risk warning is issued through the legal in-vehicle UI interface, wherein the warning includes an audible warning and UI interface information;

[0115] If the collision risk judgment result is that the collision risk is high, the corresponding drive test data is saved in the too-close-to-obstacle scenario data set.

[0116] By implementing the above method, driving data of the main vehicle during driving, target obstacles in the surrounding environment, and position information of the target obstacles are obtained, and a predicted driving trajectory of the main vehicle in the next time period in the current driving direction is calculated based on the driving data. A first distance of the predicted driving trajectory relative to the target obstacle and a second distance of the main vehicle relative to the target obstacle are calculated based on the driving data, the position information of the target obstacle, and the predicted driving trajectory. Based on the first distance and the second distance, it is determined whether the scene corresponding to the drive test data is an obstacle avoidance scene. In the above, by collecting and analyzing data during the main vehicle's driving process, and processing and comparing obstacle position information and road edge position information, target obstacles are effectively screened out. Based on the close distance detection based on the main vehicle position information and the target obstacle position information, and the close distance detection based on the predicted trajectory route and the target obstacle position information, effective detection of collision risk is achieved, thereby solving the problems of the existing technology of the inability to accurately screen obstacles and the low accuracy of obstacle avoidance scene detection.

[0117] See also Figure 3 The third embodiment of the obstacle avoidance scene detection method in the embodiment of the present invention is implemented by the following steps:

[0118] 301. Obtaining drive test data;

[0119] In this step, the road test data includes at least the position information and speed information corresponding to all test vehicles, and at least the position information and speed information corresponding to all detectable objects.

[0120] 302. Constructing a semantic map based on drive test data;

[0121] This step can be implemented in the following ways:

[0122] Parsing the drive test data, saving the position information and speed information corresponding to all the test vehicles and all the detected objects in the form of a vector diagram, and obtaining an original vector diagram;

[0123] Extracting feature information from the original vector image through a convolutional neural network and converting it into a corresponding feature vector;

[0124] Performing a weighted operation on the feature vector through a preset neural network classifier model and outputting a corresponding weighted vector;

[0125] Inputting the weighted vectors into an association probability solving model, and outputting probability values ​​of each weighted vector belonging to each spatial region;

[0126] Determining the category corresponding to each vector in the original vector graph according to the probability value;

[0127] A semantic map is constructed based on the vector type.

[0128] In practical applications, the process of constructing a semantic map based on drive test data further includes:

[0129] Input the semantic map information corresponding to the current moment into the preset prediction trajectory model to obtain the predicted trajectory of the test vehicle in the future;

[0130] The predicted trajectory of the test vehicle in the future is embedded into the semantic map.

[0131] 303. Parse the semantic map, select a target vehicle to be analyzed, and filter target obstacles corresponding to the target vehicle;

[0132] This step can be implemented in the following ways:

[0133] Obtaining an identifier corresponding to a target vehicle to be analyzed, using the identifier as an index, searching among all the test vehicles, selecting a matching vehicle, and obtaining the target vehicle to be analyzed;

[0134] Obtaining the position information of all detectable objects within a preset range around the target vehicle, for example, obtaining the position information of all detectable objects within a range of 100 meters around the target vehicle;

[0135] Parsing the position information of all detectable objects to obtain road surface position information and obstacle position information, for example, extracting a coordinate set of position points corresponding to the road surface and the position coordinates of other detectable objects to obtain road surface position information and obstacle position information;

[0136] Parsing the roadside position information and the obstacle position information to determine whether the obstacle is located within the road surface, for example, determining whether the position coordinates of other detectable objects belong to a set of position point coordinates corresponding to the road surface; if so, the obstacle is located within the road surface; if not, the obstacle is not located within the road surface;

[0137] If the obstacle is located within the road surface, it is marked as a target obstacle; if the obstacle is not located within the road surface, it is marked as an obstacle on the curb.

[0138] 304. Determine whether the distance between the target vehicle's future predicted trajectory and the target obstacle is too close based on the target vehicle's predicted trajectory and the target obstacle's position information;

[0139] This step can be implemented in the following ways:

[0140] Parse the semantic map to extract the predicted trajectory route points of the target vehicle and the location information of the target obstacles within a preset time period, wherein the preset time period can be set to within the next 3 seconds;

[0141] Calculating the lateral distance between the target vehicle and the obstacle and the longitudinal distance between the target vehicle and the obstacle based on the predicted trajectory route point of the target vehicle and the actual position of the obstacle, wherein the longitudinal distance between the target vehicle and the obstacle refers to the distance in the direction of the predicted trajectory route, and the lateral distance between the target vehicle and the obstacle refers to the distance on the road surface perpendicular to the direction of the predicted trajectory route;

[0142] Obtaining the size information of the target vehicle, calculating the distance from the rear bumper of the target vehicle to the center of the target vehicle, the distance from the front bumper of the target vehicle to the center of the target vehicle, and the lateral width of the target vehicle;

[0143] A trajectory distance threshold condition is constructed based on the distance from the rear bumper of the target vehicle to the center of the target vehicle, the distance from the front bumper of the target vehicle to the center of the target vehicle, and the lateral width of the target vehicle, and it is judged whether the longitudinal distance between the target vehicle and the obstacle and the lateral distance between the target vehicle and the obstacle meet the trajectory distance threshold condition, for example; if the longitudinal distance between the target vehicle and the obstacle is recorded as D1, the lateral distance between the target vehicle and the obstacle is recorded as D2, the distance from the front bumper of the target vehicle to the center of the target vehicle is recorded as d1, the distance from the rear bumper of the target vehicle to the center of the target vehicle is recorded as d2, and the lateral width of the target vehicle is recorded as d3, then when D1 and D2 simultaneously satisfy D1∈(d2+0.8, d1+0.8) and D2<(0.5*d3)+0.3, it is judged that the trajectory distance threshold condition is met;

[0144] If the trajectory distance threshold condition is met, it is determined that the distance between the target vehicle's future predicted trajectory and the target obstacle is too close, and the duration for which the trajectory distance threshold condition is met is recorded as a first duration;

[0145] If the trajectory distance threshold condition is not met, it is determined that the distance between the target vehicle's future predicted trajectory and the target obstacle is not too close.

[0146] 305. Determine whether the target vehicle is too close to the target obstacle based on the target vehicle's position information and the target obstacle's position information;

[0147] This step can be implemented in the following ways:

[0148] Parse the semantic map to obtain the location information of the target vehicle and the location information of the target obstacle;

[0149] Calculating a relative position vector and a relative velocity vector between the target vehicle and the target obstacle based on the position information of the target vehicle and the position information of the target obstacle;

[0150] Determining whether the modulus of the relative velocity vector is 0; if so, determining whether the modulus of the relative position vector is less than a preset stationary distance threshold; if so, determining that the target vehicle is too close to the target obstacle; and if not, determining that the target vehicle is not too close to the target obstacle;

[0151] If the modulus of the relative velocity vector is not 0, calculating the distance between the target vehicle and the target obstacle after a unit time based on the relative position vector and the relative velocity vector;

[0152] Determine whether the distance between the target vehicle and the target obstacle after the unit time is less than a preset movement distance threshold; if so, determine that the target vehicle is too close to the target obstacle, and record the duration of the target vehicle being too close to the target obstacle as a second duration;

[0153] If it is not less than, it is determined that the target vehicle is not too close to the target obstacle.

[0154] 306. Determine whether the corresponding scene is an obstacle avoidance scene based on the duration of the close distance;

[0155] Determine whether the first duration is greater than a preset first time threshold, for example, the preset first time threshold may be set to 0.5 seconds;

[0156] Determine whether the second duration is greater than a preset second time threshold, for example, the preset second time threshold may be set to 0.5 seconds;

[0157] If both the first duration and the second duration are greater than corresponding time thresholds, the corresponding scene is determined to be an obstacle avoidance scene; otherwise, the corresponding scene is determined not to be an obstacle avoidance scene.

[0158] By implementing the above method, driving data of the main vehicle during driving, target obstacles in the surrounding environment, and position information of the target obstacles are obtained, and a predicted driving trajectory of the main vehicle in the next time period in the current driving direction is calculated based on the driving data. A first duration during which the predicted driving trajectory is too close to the target obstacle and a second duration during which the main vehicle is too close to the target obstacle are calculated based on the driving data, the position information of the target obstacle, and the predicted driving trajectory. Based on the first duration and the second duration, it is determined whether the corresponding scene is an obstacle avoidance scene, thereby effectively detecting obstacle avoidance scenes. By selecting a target vehicle and determining whether obstacles within a corresponding range are located within the road surface, the target obstacles are effectively screened out, thereby solving the problems of the existing technology of being unable to accurately screen obstacles and having a low accuracy rate in obstacle avoidance scene detection.

[0159] The above describes the detection method of the obstacle avoidance scene in the embodiment of the present invention. The following describes the detection device of the obstacle avoidance scene in the embodiment of the present invention. Please refer to Figure 4 An embodiment of a device for detecting an obstacle avoidance scene in an embodiment of the present invention includes:

[0160] An acquisition module 401 is configured to acquire drive test data of the host vehicle during driving, wherein the drive test data includes driving data, target obstacles in the surrounding environment, and location information of the target obstacles;

[0161] A prediction module 402 is configured to calculate a predicted driving trajectory of the host vehicle within a next time period in the current driving direction based on the driving data;

[0162] A first calculation module 403 is configured to calculate a first distance between the predicted driving trajectory and the target obstacle based on the position information of the target obstacle and the predicted driving trajectory;

[0163] A second calculation module 404 is configured to calculate a second distance between the host vehicle and the target obstacle based on the position information of the target obstacle and the driving data;

[0164] The judgment module 405 is configured to judge whether the first distance and the second distance satisfy a preset obstacle avoidance scenario judgment condition, and determine whether the scenario corresponding to the drive test data is an obstacle avoidance scenario based on the judgment result.

[0165] By implementing the above-mentioned device, by obtaining driving data of the main vehicle during driving, target obstacles in the surrounding environment, and position information of the target obstacles, a predicted driving trajectory of the main vehicle in the next time period in the current driving direction is calculated based on the driving data. Based on the driving data, the position information of the target obstacle, and the predicted driving trajectory, a first distance of the predicted driving trajectory relative to the target obstacle and a second distance of the main vehicle relative to the target obstacle are calculated. Based on the first distance and the second distance, it is determined whether the scene corresponding to the drive test data is an obstacle avoidance scene. This solves the problems existing in the prior art of the inability to accurately screen obstacles and the low accuracy of obstacle avoidance scene detection.

[0166] See also Figure 5 Another embodiment of the obstacle avoidance scene detection device in the embodiment of the present invention includes:

[0167] An acquisition module 401 is configured to acquire drive test data of the host vehicle during driving, wherein the drive test data includes driving data, target obstacles in the surrounding environment, and location information of the target obstacles;

[0168] A prediction module 402 is configured to calculate a predicted driving trajectory of the host vehicle within a next time period in the current driving direction based on the driving data;

[0169] A first calculation module 403 is configured to calculate a first distance between the predicted driving trajectory and the target obstacle based on the position information of the target obstacle and the predicted driving trajectory;

[0170] A second calculation module 404 is configured to calculate a second distance between the host vehicle and the target obstacle based on the position information of the target obstacle and the driving data;

[0171] A judgment module 405 is configured to judge whether the first distance and the second distance satisfy a preset obstacle avoidance scenario judgment condition, and determine whether the scenario corresponding to the drive test data is an obstacle avoidance scenario based on the judgment result;

[0172] A second determination module 406 is configured to, after determining that at least one of the longitudinal distance, the lateral distance, and the second distance is satisfied, detect whether the target obstacle in the next frame of drive test data of the host vehicle satisfies the obstacle avoidance scenario determination condition; if so, determine that the scenario corresponding to the drive test data is an obstacle avoidance scenario; if not, determine that the scenario corresponding to the drive test data is a non-obstacle avoidance scenario;

[0173] In this embodiment, the acquisition module 401 includes:

[0174] A data acquisition unit 4011 is configured to acquire a semantic map and driving data corresponding to the host vehicle, wherein the driving data includes the current location information, driving speed, and driving direction of the host vehicle;

[0175] A semantic parsing unit 4012 is configured to perform semantic parsing on the semantic map to identify all obstacles therein;

[0176] The first calculating unit 4013 is configured to calculate a third distance between each obstacle and both sides of the road, and to filter out obstacles that are not within the curb range of the road from all the obstacles based on the third distance to obtain a target obstacle;

[0177] A second calculation unit 4014 is configured to calculate the coordinates of the target obstacle in the semantic map to obtain position information of the target obstacle;

[0178] In this embodiment, the prediction module 402 includes:

[0179] A data input unit 4021 is used to input the driving data into a pre-trained neural network model, wherein the neural network model includes an input layer, an embedding layer, a hidden layer, and an output layer;

[0180] An input layer unit 4022, configured to convert the driving data into a trajectory vector matrix through the input layer and output the matrix to the embedding layer;

[0181] An embedding layer unit 4023, configured to convert the trajectory vector matrix into a corresponding embedding expression matrix through the embedding layer, and output the matrix to the hidden layer;

[0182] The hidden layer unit 4024 is used to calculate the label vector corresponding to the embedded expression matrix through the hidden layer and output it to the output layer;

[0183] The output layer unit 4025 is configured to perform classification probability calculation on the embedded expression matrix and its corresponding label vector through the output layer, and output a prediction result;

[0184] A trajectory generating unit 4026 is configured to embed the prediction result into the semantic map in a graphical manner to obtain a predicted driving trajectory of the host vehicle within a certain time period in the future;

[0185] In this embodiment, the first calculation module 403 includes:

[0186] The third calculation unit 4031 is used to calculate the point position information of the host vehicle at the end point of the next time period according to the driving speed of the host vehicle and the predicted driving trajectory;

[0187] a fourth calculating unit 4032 for calculating, based on the position information of the target obstacle and the point position information, a longitudinal distance and a lateral distance of the host vehicle relative to the target obstacle at the end point of the next time period;

[0188] In this embodiment, the second calculation module 404 includes:

[0189] a fifth calculating unit 4041, configured to calculate a relative distance between the host vehicle and the target obstacle based on the current position information and the position information of the target obstacle;

[0190] A stationary determination unit 4042 is configured to identify whether the target obstacle is stationary;

[0191] a sixth calculation unit 4043 configured to calculate, when the target obstacle is stationary, a first displacement of the host vehicle in a next unit time based on the driving speed; and calculate a second distance of the host vehicle relative to the target obstacle based on the first displacement and the relative distance;

[0192] a seventh calculation unit 4044, configured to obtain a moving speed of the target obstacle if the target obstacle is not stationary, and calculate a second displacement of the host vehicle in a next unit time based on the driving speed and the moving speed;

[0193] an eighth calculating unit 4045, configured to calculate a second distance of the host vehicle relative to the target obstacle based on the second displacement and the relative distance;

[0194] In this embodiment, the judgment module 405 includes:

[0195] The first judgment unit 4051 is used to judge whether the longitudinal distance and the lateral distance meet the trajectory obstacle avoidance condition;

[0196] The second judging unit 4051 is configured to judge whether the second distance satisfies the position obstacle avoidance condition;

[0197] A first determining unit 4053 is configured to determine that the scene corresponding to the drive test data is an obstacle avoidance scene when at least one of the longitudinal distance, the lateral distance, and the second distance is satisfied;

[0198] The second determining unit 4054 is configured to determine that the scene corresponding to the drive test data is a non-obstacle avoidance scene when the longitudinal distance, the lateral distance, and the second distance are all unsatisfactory.

[0199] By implementing the above-mentioned device, by obtaining driving data of the main vehicle during driving, target obstacles in the surrounding environment, and position information of the target obstacles, a predicted driving trajectory of the main vehicle in the next time period in the current driving direction is calculated based on the driving data. Based on the driving data, the position information of the target obstacle, and the predicted driving trajectory, a first distance of the predicted driving trajectory relative to the target obstacle and a second distance of the main vehicle relative to the target obstacle are calculated. Based on the first distance and the second distance, it is determined whether the scene corresponding to the drive test data is an obstacle avoidance scene. This solves the problems existing in the prior art of the inability to accurately screen obstacles and the low accuracy of obstacle avoidance scene detection.

[0200] See also Figure 6 , an embodiment of the computer device in the embodiment of the present invention is described in detail below from the perspective of hardware processing.

[0201] Figure 6 6 is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. The computer device 600 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 610 (for example, one or more processors) and a memory 620, and one or more storage media 630 (for example, one or more mass storage devices) storing application programs 633 or data 632. Among them, the memory 620 and the storage medium 630 can be temporary storage or permanent storage. The program stored in the storage medium 630 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the computer device 600. Furthermore, the processor 610 can be configured to communicate with the storage medium 630 to execute a series of instruction operations in the storage medium 630 on the computer device 600.

[0202] The computer device 600 may further include one or more power supplies 640, one or more wired or wireless network interfaces 650, one or more input and output interfaces 660, and / or one or more operating systems 631, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 6 The computer device structure shown does not constitute a limitation on the computer device provided in this application, and may include more or fewer components than shown in the figure, or combine certain components, or arrange components differently.

[0203] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions, which, when executed on a computer, cause the computer to execute the steps of the above-mentioned obstacle avoidance scenario detection method.

[0204] In practical applications, the above-mentioned methods can be implemented based on artificial intelligence technology, where artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results. Specifically, they can be executed based on a server, which can be a standalone server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0205] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0206] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0207] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting an obstacle avoidance scene, characterized in that: The obstacle avoidance scene detection method includes: Acquire the drive test data of the main vehicle during driving, wherein the drive test data includes driving data, target obstacles in the surrounding environment and position information of the target obstacles; Calculating a predicted driving trajectory of the host vehicle in the next time period in the current driving direction based on the driving data; Calculating a first distance of the predicted driving trajectory relative to the target obstacle based on the position information of the target obstacle and the predicted driving trajectory; Calculating a second distance of the host vehicle relative to the target obstacle according to the position information of the target obstacle and the driving data; Determining whether the first distance and the second distance meet a preset obstacle avoidance scenario determination condition, and determining whether the scenario corresponding to the drive test data is an obstacle avoidance scenario based on a result of the determination; According to the road test data, determine and output the value of the distance risk coefficient corresponding to the host vehicle and the target obstacle, wherein the distance risk coefficient includes at least two values ​​0 and 1, 0 indicates that the distance between the host vehicle and the target obstacle is far and the risk is low, and 1 indicates that the distance between the host vehicle and the target obstacle is too close and the risk is high; According to the predicted driving trajectory of the main vehicle and the position information of the target obstacle, the value of the predicted trajectory distance risk coefficient corresponding to the main vehicle and the target obstacle is determined and output, wherein the predicted trajectory distance risk coefficient includes at least two values ​​0 and 1, 0 indicates that the distance between the predicted driving trajectory of the main vehicle and the target obstacle is far and the risk is low, and 1 indicates that the distance between the predicted driving trajectory of the main vehicle and the target obstacle is too close and the risk is high; Determine whether the distance risk coefficient is 1, if so, determine that the collision risk determination result is a high risk of collision; if not, determine whether the predicted trajectory distance risk coefficient is 1, if so, determine that the collision risk determination result is a high risk of collision; if not, determine that the collision risk determination result is a low risk of collision; The collision risk determination result is outputted.

2. The obstacle avoidance scene detection method according to claim 1, characterized in that: The step of obtaining the road test data of the main vehicle during driving includes: Acquire a semantic map and driving data corresponding to the main vehicle, wherein the driving data includes current position information, driving speed and driving direction of the main vehicle; Performing semantic analysis on the semantic map to identify all obstacles therein; Calculating a third distance between each obstacle and both sides of the road, and selecting obstacles that are not within the curb range of the road from all the obstacles based on the third distance to obtain a target obstacle; The coordinates of the target obstacle in the semantic map are calculated to obtain the position information of the target obstacle.

3. The obstacle avoidance scene detection method according to claim 2, characterized in that: The first distance includes a longitudinal distance and a lateral distance; and calculating the first distance of the predicted driving trajectory relative to the target obstacle according to the position information of the target obstacle and the predicted driving trajectory includes: Calculating the point position information of the main vehicle at the end point of the next time period according to the driving speed of the main vehicle and the predicted driving trajectory; The longitudinal distance and the lateral distance of the host vehicle relative to the target obstacle at the end point of the next time period are calculated according to the position information of the target obstacle and the point position information.

4. The obstacle avoidance scene detection method according to claim 3, characterized in that: The calculating, based on the position information of the target obstacle and the driving data, a second distance of the host vehicle relative to the target obstacle comprises: Calculating the relative distance between the host vehicle and the target obstacle according to the current position information and the position information of the target obstacle; Identify whether the target obstacle is in a stationary state; If the target obstacle is stationary, a first displacement of the host vehicle at the next unit time is calculated based on the driving speed; a second distance of the host vehicle relative to the target obstacle is calculated based on the first displacement and the relative distance; If the target obstacle is not stationary, the moving speed of the target obstacle is obtained, and based on the driving speed and the moving speed, a second displacement of the main vehicle at the next unit time is calculated; and according to the second displacement and the relative distance, a second distance of the main vehicle relative to the target obstacle is calculated.

5. The obstacle avoidance scene detection method according to claim 4, characterized in that: The preset obstacle avoidance scene judgment condition includes a trajectory obstacle avoidance condition and a position obstacle avoidance condition; the judging whether the first distance and the second distance meet the preset obstacle avoidance scene judgment condition, obtaining a judgment result, and determining whether the scene corresponding to the drive test data is an obstacle avoidance scene based on the judgment result, including: Determining whether the longitudinal distance and the lateral distance meet the trajectory obstacle avoidance condition; Determining whether the second distance satisfies the position obstacle avoidance condition; If at least one of the longitudinal distance, the lateral distance and the second distance satisfies the condition, determining that the scene corresponding to the drive test data is an obstacle avoidance scene; If the longitudinal distance, the lateral distance and the second distance are all unsatisfactory, it is determined that the scene corresponding to the drive test data is a non-obstacle avoidance scene.

6. The obstacle avoidance scene detection method according to claim 5, characterized in that: After determining that at least one of the longitudinal distance, the lateral distance and the second distance is satisfied, the method further includes: Detecting whether the target obstacle in the next frame of road test data of the host vehicle satisfies the obstacle avoidance scene judgment condition; If yes, determining that the scene corresponding to the drive test data is an obstacle avoidance scene; If not, it is determined that the scene corresponding to the drive test data is a non-obstacle avoidance scene.

7. The obstacle avoidance scene detection method according to claim 2, characterized in that: The step of calculating the predicted driving trajectory of the host vehicle in the next time period in the current driving direction based on the driving data includes: Inputting the driving data into a pre-trained neural network model, wherein the neural network model comprises an input layer, an embedding layer, a hidden layer and an output layer; Converting the driving data into a trajectory vector matrix through the input layer and outputting it to the embedding layer; Converting the trajectory vector matrix into a corresponding embedding expression matrix through the embedding layer and outputting it to the hidden layer; Calculate the label vector corresponding to the embedded expression matrix through the hidden layer, and output it to the output layer; Calling a classification algorithm through the output layer to calculate the classification probability of the embedded expression matrix and its corresponding label vector, and outputting a prediction result; The prediction result is embedded into the semantic map in a graphical manner to obtain the predicted driving trajectory of the main vehicle in a future time period.

8. A detection device for obstacle avoidance scene, characterized in that: The obstacle avoidance scene detection device includes: An acquisition module, used to acquire the drive test data of the main vehicle during driving, wherein the drive test data includes driving data, target obstacles in the surrounding environment and location information of the target obstacles; A prediction module, used to calculate a predicted driving trajectory of the host vehicle in the next time period in the current driving direction based on the driving data; A first calculation module, configured to calculate a first distance of the predicted driving trajectory relative to the target obstacle according to the position information of the target obstacle and the predicted driving trajectory; A second calculation module, configured to calculate a second distance of the host vehicle relative to the target obstacle according to the position information of the target obstacle and the driving data; A judgment module, used to judge whether the first distance and the second distance meet a preset obstacle avoidance scene judgment condition, and determine whether the scene corresponding to the drive test data is an obstacle avoidance scene based on the judgment result; The device also includes a collision risk judgment module, which is used to determine the value of the distance risk coefficient corresponding to the main vehicle and the target obstacle based on the road test data and output it, wherein the distance risk coefficient includes at least two values ​​0 and 1, 0 indicates that the distance between the main vehicle and the target obstacle is far and the risk is low, and 1 indicates that the distance between the main vehicle and the target obstacle is too close and the risk is high; according to the predicted driving trajectory of the main vehicle and the position information of the target obstacle, determine the value of the predicted trajectory distance risk coefficient corresponding to the main vehicle and the target obstacle and output it, wherein the predicted trajectory distance risk The coefficient includes at least two values ​​0 and 1, 0 indicates that the distance between the predicted driving trajectory of the main vehicle and the target obstacle is far and the risk is low, 1 indicates that the distance between the predicted driving trajectory of the main vehicle and the target obstacle is too close and the risk is high; judge whether the distance risk coefficient is 1, if so, determine that the collision risk judgment result is a high risk of collision; if not, judge whether the predicted trajectory distance risk coefficient is 1, if so, determine that the collision risk judgment result is a high risk of collision; if not, determine that the collision risk judgment result is a low risk of collision; output the collision risk judgment result.

9. A computer device, characterized in that: include: A memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via a line; The at least one processor calls the instructions in the memory so that the computer device executes each step of the obstacle avoidance scene detection method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the obstacle avoidance scene detection method as claimed in any one of claims 1 to 7 is implemented.

Images