An intelligent mapping method, system and device for outdoor parking lots with synchronous positioning

Abstract

The present invention discloses an intelligent mapping method, system and device for outdoor parking lots with synchronous positioning, which relates to the technical field of mapping. The method includes: after a vehicle enters the parking lot, the position information of the vehicle is obtained in real time through a positioning device, and the behavior monitoring data of the vehicle is obtained through an in-vehicle electronic system, wherein the behavior monitoring data includes at least one of in-vehicle sensor data and driving operation data; determining the driving trajectory of the vehicle according to the position information, and determining the parking position of the vehicle according to the driving trajectory; based on a reference parking space, determining the parking attitude of the vehicle according to the behavior monitoring data and the driving trajectory; determining the parking line data of the vehicle at the parking position according to the parking attitude, and updating the parking space mapping map of the parking lot according to the parking line data. The present invention has the effect of improving the intelligent level of parking lot mapping.

Description

Technical Field

[0001] The present invention relates to the technical field of surveying and mapping, and in particular, to an intelligent surveying and mapping method, system and device for outdoor parking lots with synchronous positioning. Background Art

[0002] With the acceleration of the urbanization process, the increase in urban population has led to a sharp increase in the number of motor vehicles, and the problem of parking difficulty has become increasingly prominent. In this context, traditional parking lot management methods have become increasingly ineffective. Many parking lots still rely on manual management and static maps, and cannot reflect the usage of parking spaces in a timely manner. Existing surveying and mapping technologies mostly rely on manual measurement and simple electronic systems, lacking real-time performance and automation, resulting in a lag in the update of parking space information. When users are looking for parking spaces, they often face a dilemma of not knowing what to do.

[0003] With the rapid development of technologies such as the Internet of Things, cloud computing, and big data analysis, some modern parking management systems have begun to use sensors and basic positioning technologies. The intelligent transformation of parking lot surveying and mapping and management has become possible. However, at present, most parking lots have not fully realized this transformation, the popularity of intelligent configurations and technologies is relatively low, and many systems still rely on static surveying and mapping information, lacking rapid response and dynamic management of the usage of parking spaces. Therefore, the current intelligent level of parking lot surveying and mapping is relatively low, resulting in low management efficiency of parking lots. Summary of the Invention

[0004] In order to improve the intelligent level of parking lot surveying and mapping, the present invention provides an intelligent surveying and mapping method, system and device for outdoor parking lots with synchronous positioning.

[0005] In a first aspect, the present invention provides an intelligent surveying and mapping method for outdoor parking lots with synchronous positioning, adopting the following technical solutions:

[0006] An intelligent surveying and mapping method for outdoor parking lots with synchronous positioning, comprising:

[0007] After a vehicle enters the parking lot, the position information of the vehicle is obtained in real time through a positioning device, and the behavior monitoring data of the vehicle is obtained through an in-vehicle electronic system, wherein the behavior monitoring data includes at least one of in-vehicle sensor data and driving operation data;

[0008] The driving trajectory of the vehicle is determined according to the position information, and the parking position of the vehicle is determined according to the driving trajectory;

[0009] Based on a reference parking space, the parking attitude of the vehicle is determined according to the behavior monitoring data and the driving trajectory, wherein the parking attitude includes the parking angle;

[0010] Determine the parking line data of the vehicle at the parking position according to the parking attitude, and update the parking space mapping map of the parking lot according to the parking line data.

[0011] By adopting the above technical solution, after the vehicle enters the parking lot, the position information of the vehicle is obtained in real time through the positioning device, and the behavior monitoring data of the vehicle is obtained through the in-vehicle electronic system, wherein the behavior monitoring data includes at least one of in-vehicle sensor data and driving operation data. Then, the driving trajectory of the vehicle is determined according to the position information, and the parking position of the vehicle is determined according to the driving trajectory. Then, based on the reference parking space, the parking attitude of the vehicle is determined according to the behavior monitoring data and the driving trajectory, wherein the parking attitude includes the parking angle. Finally, the parking line data of the vehicle at the parking position is determined according to the parking attitude, and the parking space mapping map of the parking lot is updated according to the parking line data; through the above method, the parking space can be mapped by the data collected by the vehicle itself driving into the parking lot, without additionally installing a data collection device in the parking lot, thereby improving the real-time performance of the parking lot mapping, and improving the intelligent level of the parking lot mapping and the management efficiency of the parking lot. In addition, the monitoring of the vehicles in the parking lot is realized, and the safety during the parking process is improved.

[0012] Optionally, the step of determining the driving trajectory of the vehicle according to the position information includes:

[0013] Based on the Kalman filter algorithm, generate Kalman position information according to the position information;

[0014] Perform trajectory interpolation on the Kalman position information according to the Kalman position information to obtain the driving trajectory of the vehicle.

[0015] By adopting the above technical solution, in order to determine the driving trajectory of the vehicle according to the position information, first, based on the Kalman filter algorithm, generate Kalman position information according to the position information, and then perform trajectory interpolation on the Kalman position information according to the Kalman position information to obtain the driving trajectory of the vehicle.

[0016] Optionally, the step of determining the parking position of the vehicle according to the driving trajectory includes:

[0017] Obtain the timestamp data corresponding to the driving trajectory according to the driving trajectory, and create an initial window according to the preset time length, the driving trajectory and the timestamp data;

[0018] Slide the initial window according to the preset time sliding unit to obtain at least one intermediate window;

[0019] For each of the intermediate windows, determine whether the distances between all adjacent position points within the intermediate window are all less than a preset distance, and determine whether the residence time of multiple position points at the same position exceeds a preset time. If both conditions are met, determine the parking data of the vehicle according to the intermediate window, where the parking data includes the parking position, the parking time period, and the parking duration.

[0020] By adopting the above technical solution, in order to determine the parking position of the vehicle according to the driving trajectory, obtain the timestamp data corresponding to the driving trajectory according to the driving trajectory, create an initial window according to the preset time length, the driving trajectory, and the timestamp data, then slide the initial window according to the preset time sliding unit to obtain at least one intermediate window. Finally, for each intermediate window, determine whether the distances between all adjacent position points within the intermediate window are all less than a preset distance, and determine whether the residence time of multiple position points at the same position exceeds a preset time. If both conditions are met, determine the parking data of the vehicle according to the intermediate window, where the parking data includes the parking position, the parking time period, and the parking duration.

[0021] Optionally, the step of determining the parking attitude of the vehicle based on a reference parking space and according to the behavior monitoring data and the driving trajectory includes:

[0022] Obtain the parking line data of the reference parking space based on the reference parking space, where the parking line data includes the direction data of the parking line;

[0023] Obtain model training data, and divide the model training data into a training set and a test set according to a preset ratio, where the model training data includes the parking line data, historical behavior detection data, and historical driving trajectories;

[0024] Set the hyperparameters of a pre-constructed deep learning model according to the random network search algorithm, and use the root mean square error RMSE and the coefficient of determination R 2 as evaluation indicators;

[0025] Train the deep learning model according to the training set to obtain a trained deep learning model;

[0026] Test the trained deep learning model according to the test set, and determine whether the error is within a preset range according to the evaluation indicators. If so, use the trained deep learning model as a parking attitude prediction model;

[0027] Based on the parking attitude prediction model, generate the parking attitude of the vehicle according to the behavior monitoring data and the driving trajectory.

[0028] By adopting the above technical solutions, in order to determine the parking attitude of the vehicle, parking line data of the reference parking space is obtained based on the reference parking space, where the parking line data includes the direction data of the parking line. Then, model training data is obtained and divided into a training set and a test set according to a preset ratio. The model training data includes parking line data, historical behavior detection data, and historical driving trajectories. Then, hyperparameters of a pre-constructed deep learning model are set according to the random network search algorithm, and the root mean square error RMSE and the coefficient of determination R 2 are used as evaluation indicators. Then, the deep learning model is trained according to the training set to obtain a trained deep learning model. Then, the trained deep learning model is tested according to the test set, and it is judged whether the error is within a preset range according to the evaluation indicators. If the error is within the preset range, the trained deep learning model is used as a parking attitude prediction model. Finally, based on the parking attitude prediction model, according to the behavior monitoring data and the driving trajectory, the parking attitude of the vehicle is generated.

[0029] Optionally, the parking attitude prediction model includes an input layer, a first feature extraction layer, a second feature extraction layer, a splicing layer, and an output layer. The step of generating the parking attitude of the vehicle according to the behavior monitoring data and the driving trajectory based on the parking attitude prediction model includes:

[0030] Through the input layer, a first data feature vector is generated according to the behavior monitoring data;

[0031] Through the input layer, a second data feature vector is generated according to the driving trajectory;

[0032] Through the first feature extraction layer, a third data feature vector is generated according to the first data feature vector;

[0033] Through the second feature extraction layer, a fourth data feature vector is generated according to the second data feature vector;

[0034] Through the splicing layer, the third data feature vector and the fourth data feature vector are spliced to obtain a fifth data feature vector;

[0035] Through the output layer, an output vector is generated according to the fifth data feature vector. The output vector includes a first element, a second element, a third element, and a fourth element. The first element is used to represent the deflection angle between the front line of the vehicle and the first side of the reference parking space, the second element is used to represent the deflection angle between the left line of the vehicle and the second side of the reference parking space, the third element is used to represent the deflection angle between the rear side line of the vehicle and the third side of the reference parking space, and the fourth element is used to represent the deflection angle between the right line of the vehicle and the fourth side of the reference parking space;

[0036] Generate the parking attitude of the vehicle based on the first element, the second element, the third element, and the fourth element.

[0037] By adopting the above technical solution, in order to generate the parking attitude of the vehicle, through the input layer, according to the behavior monitoring data, a first data feature vector is generated, and at the same time, through the input layer, according to the driving trajectory, a second data feature vector is generated. Then, through the first feature extraction layer, according to the first data feature vector, a third data feature vector is generated. Then, through the second feature extraction layer, according to the second data feature vector, a fourth data feature vector is generated. Then, through the splicing layer, the third data feature vector and the fourth data feature vector are spliced to obtain a fifth data feature vector. Then, through the output layer, according to the fifth data feature vector, an output vector is generated, where the output vector includes the first element, the second element, the third element, and the fourth element. The first element is used to represent the deflection angle between the front edge line of the vehicle and the first side of the reference parking space, the second element is used to represent the deflection angle between the left side line of the vehicle and the second side of the reference parking space, the third element is used to represent the deflection angle between the rear side line of the vehicle and the third side of the reference parking space, and the fourth element is used to represent the deflection angle between the right side line of the vehicle and the fourth side of the reference parking space. Finally, based on the first element, the second element, the third element, and the fourth element, the parking attitude of the vehicle is generated.

[0038] Optionally, the first feature extraction layer includes a first fully connected layer, a second fully connected layer, and a third fully connected layer. The neuron number ratio of the first fully connected layer, the second fully connected layer, and the third fully connected layer is 5:2:1. The step of generating the third data feature vector according to the first data feature vector through the first feature extraction layer includes:

[0039] Input the first data feature vector into the first fully connected layer to obtain a first intermediate feature vector;

[0040] Input the first intermediate feature vector into the second fully connected layer to obtain a second intermediate feature vector;

[0041] Input the second intermediate feature vector into the third fully connected layer to obtain a third data feature vector.

[0042] By adopting the above technical solution, in order to generate the third data feature vector, first input the first data feature vector into the first fully connected layer to obtain a first intermediate feature vector, then input the first intermediate feature vector into the second fully connected layer to obtain a second intermediate feature vector, and finally input the second intermediate feature vector into the third fully connected layer to obtain a third data feature vector.

[0043] Optionally, the step of determining the parking line data of the vehicle at the parking position according to the parking posture includes:

[0044] Obtain the dimension information of the vehicle, where the dimension information includes length and width;

[0045] Based on the standard data of the parking spaces in the parking lot and according to the dimension information, determine the parking space specifications of the vehicle;

[0046] Based on the parking posture of the vehicle and according to the parking space specifications, determine the parking line data of the vehicle at the parking position.

[0047] By adopting the above technical solution, in order to determine the parking line data of the vehicle at the parking position according to the parking posture, obtain the dimension information of the vehicle, where the dimension information includes length and width, then based on the standard data of the parking spaces in the parking lot and according to the dimension information, determine the parking space specifications of the vehicle, and finally based on the parking posture of the vehicle and according to the parking space specifications, determine the parking line data of the vehicle at the parking position.

[0048] In a second aspect, the present invention also provides an intelligent mapping system for outdoor parking lots with synchronous positioning, adopting the following technical solution:

[0049] An intelligent mapping system for outdoor parking lots with synchronous positioning includes:

[0050] A data acquisition module, configured to, after the vehicle enters the parking lot, acquire the position information of the vehicle in real time through a positioning device, and acquire the behavior monitoring data of the vehicle through an in-vehicle electronic system, where the behavior monitoring data includes at least one of in-vehicle sensor data and driving operation data;

[0051] A position information processing module, configured to determine the driving trajectory of the vehicle according to the position information and determine the parking position of the vehicle according to the driving trajectory;

[0052] A parking posture generation module, configured to determine the parking posture of the vehicle based on a reference parking space and according to the behavior monitoring data and the driving trajectory, where the parking posture includes a parking angle;

[0053] A parking space mapping map updating module, configured to determine the parking line data of the vehicle at the parking position according to the parking posture and update the parking space mapping map of the parking lot according to the parking line data.

[0054] The position information processing module includes:

[0055] A Kalman filter sub-module, configured to generate Kalman position information based on the Kalman filter algorithm and according to the position information;

[0056] A trajectory interpolation sub-module, configured to perform trajectory interpolation on the Kalman position information according to the Kalman position information to obtain the driving trajectory of the vehicle.

[0057] In a third aspect, the present invention further provides a computer device, adopting the following technical solution:

[0058] A computer device includes a memory and a processor. A computer program that can run on the processor is stored on the memory. When the processor executes the computer program, the method described in the first aspect is implemented.

[0059] In summary, the present invention at least includes the following beneficial technical effects: After the vehicle enters the parking lot, the position information of the vehicle is obtained in real time through a positioning device, and the behavior monitoring data of the vehicle is obtained through an in-vehicle electronic system, where the behavior monitoring data includes at least one of in-vehicle sensor data and driving operation data. Then, the driving trajectory of the vehicle is determined according to the position information, and the parking position of the vehicle is determined according to the driving trajectory. Then, based on a reference parking space, the parking attitude of the vehicle is determined according to the behavior monitoring data and the driving trajectory, where the parking attitude includes the parking angle. Finally, the parking line data of the vehicle at the parking position is determined according to the parking attitude, and the parking space mapping map of the parking lot is updated according to the parking line data; through the above method, the parking spaces can be mapped by the data collected by the vehicles driving into the parking lot itself, without the need to additionally install data collection devices in the parking lot, thereby improving the real-time performance of parking lot mapping, and improving the intelligent level of parking lot mapping and the management efficiency of the parking lot. In addition, the monitoring of vehicles in the parking lot is realized, and the safety during the parking process is improved. Description of the Drawings

[0060] Figure 1 is the overall flow schematic diagram of the embodiment of the present invention.

[0061] Figure 2 is the structural schematic diagram of the system of the present invention.

[0062] Figure 3 is the structural block diagram of the computer device of the present invention. Detailed Embodiments

[0063] In order to make the objectives, technical solutions and advantages of the present invention clearer, the following is a further detailed description of the present invention in conjunction with Figures 1-3 and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0064] The embodiment of the present invention discloses an intelligent mapping method for an outdoor parking lot with synchronous positioning.

[0065] Refer toFigure 1 , an intelligent mapping method for outdoor parking lots with synchronous positioning, including:

[0066] Step S11, after the vehicle enters the parking lot, the position information of the vehicle is obtained in real time through a positioning device, and the behavior monitoring data of the vehicle is obtained through an in-vehicle electronic system.

[0067] Among them, the behavior monitoring data includes at least one of in-vehicle sensor data and driving operation data;

[0068] It should be noted that in step S11, the vehicle is monitored in real time through the positioning device and / or the in-vehicle electronic system. Usually, a GPS module or other positioning systems (such as GNSS) are installed in the vehicle to ensure accurate position information can be obtained even in the parking lot. Technologies such as base station positioning or Wi-Fi positioning can also be combined to improve the positioning accuracy; the in-vehicle electronic system is connected to in-vehicle sensors (such as accelerometers, gyroscopes, radar sensors, etc.) to collect the status information of the vehicle in real time, or monitor the operation data of the driver (i.e., driving operation data), including information such as throttle, brake, steering wheel angle, and gear. These data can be read through the OBD-II interface or the in-vehicle controller; then the monitored behavior data is uploaded to the cloud platform through the in-vehicle communication system (such as 4G / 5G, V2X communication), or the data is transmitted to the parking lot management system through Wi-Fi for subsequent processing.

[0069] Step S12, determine the driving trajectory of the vehicle according to the position information, and determine the parking position of the vehicle according to the driving trajectory.

[0070] Step S13, based on the reference parking space, determine the parking posture of the vehicle according to the behavior monitoring data and the driving trajectory.

[0071] Among them, the parking posture includes the parking angle.

[0072] It should be noted that the parking posture of the vehicle is used to represent the posture or state of the vehicle relative to the reference parking space. For example, if the parking lot is a flat ground, the parking posture may include the parking angle of the vehicle relative to the reference parking space. If the parking lot includes two flat grounds with different heights, the parking posture may include the parking height of the vehicle relative to the reference parking space.

[0073] Step S14, determine the parking line data of the vehicle at the parking position according to the parking posture, and update the parking space mapping map of the parking lot according to the parking line data.

[0074] It can be understood that at the beginning of the mapping, the parking space mapping map is usually an initial map without parking lines. During the mapping process, the parking space mapping map is a map formed by adding a part of the parking lines to the initial map.

[0075] In the above-described embodiment, after the vehicle enters the parking lot, the position information of the vehicle is obtained in real time through the positioning device, and the behavior monitoring data of the vehicle is obtained through the in-vehicle electronic system, where the behavior monitoring data includes at least one of in-vehicle sensor data and driving operation data. Then, the driving trajectory of the vehicle is determined based on the position information, and the parking position of the vehicle is determined based on the driving trajectory. Then, based on the reference parking space, the parking attitude of the vehicle is determined according to the behavior monitoring data and the driving trajectory, where the parking attitude includes the parking angle. Finally, the parking line data of the vehicle at the parking position is determined according to the parking attitude, and the parking space mapping map of the parking lot is updated according to the parking line data; through the above method, the parking space can be mapped by the data collected by the vehicle itself driving into the parking lot, without the need to additionally install data collection devices in the parking lot, thereby improving the real-time performance of parking lot mapping, and improving the intelligent level of parking lot mapping and the management efficiency of the parking lot. In addition, the monitoring of vehicles in the parking lot is realized, and the safety during the parking process is improved.

[0076] As a further embodiment of the method, the step of determining the driving trajectory of the vehicle based on the position information includes:

[0077] Step S21, based on the Kalman filter algorithm, generate Kalman position information according to the position information.

[0078] It should be noted that the Kalman filter is a recursive algorithm for estimating the state of a linear dynamic system, which is widely used in signal processing and control systems. Its main advantage is that it can effectively process measurement data containing noise and can estimate the system state. Suppose for a vehicle, our goal is to estimate its position (i.e., the predicted position) through the Kalman filter. We obtain measurement values from sensors (such as GPS), but these measurement values often contain noise. The following are the specific steps:

[0079] First, perform state prediction. Assume that the true position of the vehicle is (2, 3) at a certain moment. Through Kalman filtering, the position of the vehicle at the next moment (e.g., per second) can be predicted. Based on the previous position and speed, the predicted position (i.e., state estimation) may be (2.5, 3.5). Then, obtain the corresponding measurement value through the position information. Due to the interference of GPS signals, the measurement value we get may be (2.7, 3.2). This measurement value has some deviation from the true position, but we hope to use it to correct our predicted position. Then, the Kalman filtering algorithm will consider the predicted value (2.5, 3.5) and the measurement value (2.7, 3.2) in the update step, and calculate a new estimated position based on the confidence levels (through weighting) of the two. This can be achieved through a weighting operation. Assume that the weight given to the predicted value is 0.7 and the weight given to the measurement value is 0.3. We will calculate a weighted average: new position = 0.7×(2.5, 3.5) + 0.3×(2.7, 3.2). This can make the new position estimate closer to the predicted value while also considering the small changes brought by the measurement value. In Kalman filtering, we will maintain a covariance matrix representing uncertainty. In the update step, the covariance matrix is used to update the accuracy of the state. For example, a smaller covariance value indicates a higher confidence in the prediction. Therefore, when updating, the weight of the predicted position is relatively increased. Finally, after weighted calculation and covariance adjustment, the new estimated position of the vehicle we get may be (2.6, 3.4). This position combines the past motion prediction and the current measurement information well, minimizing the deviation introduced by measurement errors.

[0080] Step S22: Perform trajectory interpolation on the Kalman position information according to the Kalman position information to obtain the driving trajectory of the vehicle.

[0081] It can be understood that through trajectory interpolation, missing information can be filled between known data points, thus forming a continuous and smooth trajectory without jumps or unnatural changes.

[0082] In the above embodiment, in order to determine the driving trajectory of the vehicle according to the position information, first, based on the Kalman filtering algorithm, and according to the position information, generate Kalman position information. Then, perform trajectory interpolation on the Kalman position information according to the Kalman position information to obtain the driving trajectory of the vehicle.

[0083] As a further embodiment of the method, the steps of determining the parking position of the vehicle according to the driving trajectory include:

[0084] Step S31: Obtain the timestamp data corresponding to the driving trajectory according to the driving trajectory, and create an initial window according to the preset time length, driving trajectory, and timestamp data.

[0085] Step S32: Slide the initial window according to a preset time sliding unit to obtain at least one intermediate window.

[0086] Step S33: For each intermediate window, determine whether the distances between all adjacent position points within the intermediate window are all less than a preset distance, and determine whether the residence times of multiple position points at the same position exceed a preset time. If both conditions are met, determine the parking data of the vehicle according to the intermediate window.

[0087] Among them, the parking data includes the parking position, the parking time period, and the parking duration.

[0088] In the above embodiment, in order to determine the parking position of the vehicle according to the driving trajectory, obtain the timestamp data corresponding to the driving trajectory according to the driving trajectory, create an initial window according to the preset time length, the driving trajectory, and the timestamp data, then slide the initial window according to the preset time sliding unit to obtain at least one intermediate window, and finally, for each intermediate window, determine whether the distances between all adjacent position points within the intermediate window are all less than a preset distance, and determine whether the residence times of multiple position points at the same position exceed a preset time. If both conditions are met, determine the parking data of the vehicle according to the intermediate window. Among them, the parking data includes the parking position, the parking time period, and the parking duration.

[0089] As a further embodiment of the method, the steps of determining the parking attitude of the vehicle based on a reference parking space and according to behavior monitoring data and driving trajectory include:

[0090] Step S41: Obtain the parking line data of the reference parking space based on the reference parking space, where the parking line data includes the direction data of the parking line.

[0091] Step S42: Obtain model training data and divide the model training data into a training set and a test set according to a preset ratio.

[0092] Among them, the model training data includes parking line data, historical behavior detection data, and historical driving trajectory.

[0093] Step S43: Set the hyperparameters of a pre-constructed deep learning model according to the random network search algorithm, and use the root mean square error RMSE and the coefficient of determination R 2 as evaluation indicators.

[0094] It should be noted that random network search is a method for hyperparameter tuning. Compared with grid search, it randomly selects parameter combinations for model training and evaluation. Its advantages include high efficiency: when the parameter space is large, random selection is more effective than systematically traversing each parameter combination and can quickly find relatively good parameter settings; broad exploration: it can cover a wider parameter space and reduce the dependence on local optima. In deep learning models, the hyperparameters that often need to be tuned include: learning rate (such as 0.001, 0.01, 0.1), batch size (such as 16, 32, 64, 128), number of network layers and the number of neurons in each layer, activation function (such as ReLU, Sigmoid, Tanh), and regularization parameters (such as dropout rate, L2 regularization, etc.)

[0095] Step S44: Train the deep learning model according to the training set to obtain a trained deep learning model.

[0096] Step S45: Test the trained deep learning model according to the test set, and judge whether the error is within the preset range according to the evaluation index. If so, use the trained deep learning model as the parking pose prediction model.

[0097] Specifically, test the trained deep learning model according to the test set, and judge whether the error is within the preset range according to the evaluation index. If the error is within the preset range, use the trained deep learning model as the parking pose prediction model.

[0098] Step S46: Based on the parking pose prediction model, generate the parking pose of the vehicle according to the behavior monitoring data and driving trajectory.

[0099] In the above embodiment, in order to determine the parking pose of the vehicle, parking line data of the reference parking space is obtained based on the reference parking space. Among them, the parking line data includes the direction data of the parking line. Then, model training data is obtained and divided into a training set and a test set according to a preset ratio. Among them, the model training data includes parking line data, historical behavior detection data, and historical driving trajectory. Then, the hyperparameters of the pre-constructed deep learning model are set according to the random network search algorithm, and the root mean square error RMSE and the coefficient of determination R 2 are used as evaluation indexes. Then, the deep learning model is trained according to the training set to obtain a trained deep learning model. Then, the trained deep learning model is tested according to the test set, and it is judged whether the error is within the preset range according to the evaluation index. If the error is within the preset range, the trained deep learning model is used as the parking pose prediction model. Finally, based on the parking pose prediction model, the parking pose of the vehicle is generated according to the behavior monitoring data and driving trajectory.

[0100] As a further implementation of the method, the parking attitude prediction model includes an input layer, a first feature extraction layer, a second feature extraction layer, a concatenation layer, and an output layer. Based on the parking attitude prediction model, the steps of generating the parking attitude of the vehicle according to the behavior monitoring data and the driving trajectory include:

[0101] Step S51: Through the input layer, generate a first data feature vector according to the behavior monitoring data.

[0102] Step S52: Through the input layer, generate a second data feature vector according to the driving trajectory.

[0103] It should be noted that in a deep learning model, the input layer is the first layer of the model, responsible for receiving external data and converting it into a format suitable for model processing. The input can be a feature vector, an image, text, etc., depending on the specific task. The input layer usually needs to specify the shape and dimension of the data so that the subsequent network layers can process it correctly.

[0104] Step S53: Through the first feature extraction layer, generate a third data feature vector according to the first data feature vector.

[0105] Step S54: Through the second feature extraction layer, generate a fourth data feature vector according to the second data feature vector.

[0106] Step S55: Through the concatenation layer, perform a concatenation process on the third data feature vector and the fourth data feature vector to obtain a fifth data feature vector.

[0107] It should be noted that the concatenation layer is a layer commonly used in deep learning, especially in convolutional neural networks (CNNs). Its main function is to connect multiple input tensors along a specified dimension to form a new tensor. This concatenation operation can help the model fuse the features of different layers and achieve stronger representation capabilities.

[0108] Step S56: Through the output layer, generate an output vector according to the fifth data feature vector.

[0109] Among them, the output vector includes a first element, a second element, a third element, and a fourth element. The first element is used to represent the deflection angle between the front line of the vehicle and the first side of the reference parking space. The second element is used to represent the deflection angle between the left line of the vehicle and the second side of the reference parking space. The third element is used to represent the deflection angle between the rear side line of the vehicle and the third side of the reference parking space. The fourth element is used to represent the deflection angle between the right line of the vehicle and the fourth side of the reference parking space.

[0110] It should be noted that the output layer is the last layer in the deep learning model, which is responsible for converting the features generated by the model into the final prediction results. According to different task types, the structure and activation function of the output layer will also be different.

[0111] Step S57: Generate the parking posture of the vehicle based on the first element, the second element, the third element, and the fourth element.

[0112] In the above embodiment, in order to generate the parking posture of the vehicle, through the input layer, according to the behavior monitoring data, a first data feature vector is generated, and at the same time, through the input layer, according to the driving trajectory, a second data feature vector is generated. Then, through the first feature extraction layer, according to the first data feature vector, a third data feature vector is generated. Then, through the second feature extraction layer, according to the second data feature vector, a fourth data feature vector is generated. Then, through the splicing layer, the third data feature vector and the fourth data feature vector are spliced to obtain a fifth data feature vector. Then, through the output layer, according to the fifth data feature vector, an output vector is generated, where the output vector includes the first element, the second element, the third element, and the fourth element. The first element is used to represent the deflection angle between the front line of the vehicle and the first side of the reference parking space, the second element is used to represent the deflection angle between the left line of the vehicle and the second side of the reference parking space, the third element is used to represent the deflection angle between the rear side line of the vehicle and the third side of the reference parking space, and the fourth element is used to represent the deflection angle between the right line of the vehicle and the fourth side of the reference parking space. Finally, based on the first element, the second element, the third element, and the fourth element, the parking posture of the vehicle is generated.

[0113] As a further embodiment of the method, the first feature extraction layer includes a first fully connected layer, a second fully connected layer, and a third fully connected layer. The neuron number ratio of the first fully connected layer, the second fully connected layer, and the third fully connected layer is 5:2:1. The step of generating the third data feature vector according to the first data feature vector through the first feature extraction layer includes:

[0114] Step S61: Input the first data feature vector into the first fully connected layer to obtain a first intermediate feature vector.

[0115] Step S62: Input the first intermediate feature vector into the second fully connected layer to obtain a second intermediate feature vector.

[0116] Step S63: Input the second intermediate feature vector into the third fully connected layer to obtain a third data feature vector.

[0117] In the above embodiment, in order to generate the third data feature vector, the first data feature vector is first input into the first fully connected layer to obtain the first intermediate feature vector, then the first intermediate feature vector is input into the second fully connected layer to obtain the second intermediate feature vector, and finally the second intermediate feature vector is input into the third fully connected layer to obtain the third data feature vector.

[0118] As a further embodiment of the method, the step of determining the parking line data of the vehicle at the parking position according to the parking attitude includes:

[0119] Step S71, obtaining the size information of the vehicle.

[0120] Among them, the size information includes the length and the width.

[0121] Step S72, based on the standard data of the parking spaces in the parking lot and according to the size information, determining the parking space specifications of the vehicle.

[0122] It should be noted that the standard data of the parking spaces is used to represent the parking space specifications of different types of vehicles.

[0123] Step S73, based on the parking attitude of the vehicle and according to the parking space specifications, determining the parking line data of the vehicle at the parking position.

[0124] In the above embodiment, in order to determine the parking line data of the vehicle at the parking position according to the parking attitude, the size information of the vehicle is obtained, where the size information includes the length and the width, then based on the standard data of the parking spaces in the parking lot and according to the size information, the parking space specifications of the vehicle are determined, and finally based on the parking attitude of the vehicle and according to the parking space specifications, the parking line data of the vehicle at the parking position is determined.

[0125] The embodiment of the present invention also discloses an intelligent mapping system for outdoor parking lots with simultaneous localization.

[0126] Reference Figure 2 , an intelligent mapping system for outdoor parking lots with simultaneous localization, includes:

[0127] A data acquisition module, configured to, after the vehicle enters the parking lot, obtain the position information of the vehicle in real time through a positioning device, and obtain the behavior monitoring data of the vehicle through an in-vehicle electronic system, where the behavior monitoring data includes at least one of in-vehicle sensor data and driving operation data;

[0128] A position information processing module, configured to determine the driving trajectory of the vehicle according to the position information and determine the parking position of the vehicle according to the driving trajectory;

[0129] A parking attitude generation module, configured to determine the parking attitude of the vehicle based on a reference parking space and according to the behavior monitoring data and the driving trajectory, where the parking attitude includes a parking angle;

[0130] The parking space surveying and mapping map update module is used to determine the parking line data of the vehicle at the parking position according to the parking attitude, and update the parking space surveying and mapping map of the parking lot according to the parking line data.

[0131] The position information processing module includes:

[0132] The Kalman filter sub-module is used to generate Kalman position information based on the Kalman filter algorithm and according to the position information;

[0133] The trajectory interpolation sub-module is used to perform trajectory interpolation on the Kalman position information according to the Kalman position information to obtain the driving trajectory of the vehicle.

[0134] An intelligent surveying and mapping system for outdoor parking lots with synchronous positioning according to the present invention can implement any one of the methods in an intelligent surveying and mapping method for outdoor parking lots with synchronous positioning, and the specific working process of the intelligent surveying and mapping system for outdoor parking lots with synchronous positioning according to the present invention can refer to the corresponding process in the above-mentioned intelligent surveying and mapping method for outdoor parking lots with synchronous positioning.

[0135] An embodiment of the present invention also discloses a computer device.

[0136] Reference Figure 3 , a computer device, including a memory and a processor, where a computer program that can run on the processor is stored on the memory, and when the processor executes the computer program, it implements any one of the above-mentioned intelligent surveying and mapping methods for outdoor parking lots with synchronous positioning.

[0137] The above are all preferred embodiments of the present invention. The protection scope of the present invention is not limited thereby. Any feature disclosed in this specification (including the abstract and drawings), unless specifically described, can be replaced by other equivalent or similar-purpose alternative features. That is, unless specifically described, each feature is only an example of a series of equivalent or similar features.

Claims

1. A synchronous positioning outdoor parking lot intelligent mapping method, characterized in that: include: After the vehicle enters the parking lot, the location information of the vehicle is obtained in real time through a positioning device, and the behavior monitoring data of the vehicle is obtained through an on-board electronic system, wherein the behavior monitoring data includes at least one of on-board sensor data and driving operation data; Determining a driving trajectory of the vehicle according to the position information, and determining a parking position of the vehicle according to the driving trajectory; Based on a reference parking space and according to the behavior monitoring data and the driving trajectory, determining a parking posture of the vehicle, wherein the parking posture includes a parking angle; Determining parking line data of the vehicle at the parking position according to the parking posture, and updating the parking space mapping map of the parking lot according to the parking line data; The step of determining the parking posture of the vehicle based on the reference parking space and according to the behavior monitoring data and the driving trajectory includes: Acquiring parking line data of the reference parking space based on the reference parking space, wherein the parking line data includes direction data of the parking line; Acquire model training data, and divide the model training data into a training set and a test set according to a preset ratio, wherein the model training data includes the parking line data, historical behavior detection data, and historical driving trajectory; The hyperparameters of the pre-built deep learning model were set according to the random network search algorithm, and the root mean square error RMSE and the coefficient of determination R 2 As an evaluation indicator; Training the deep learning model according to the training set to obtain a trained deep learning model; Testing the trained deep learning model according to the test set, and judging whether the error is within a preset range according to the evaluation index, and if so, using the trained deep learning model as a parking posture prediction model; Based on a parking posture prediction model, the parking posture of the vehicle is generated according to the behavior monitoring data and the driving trajectory.

2. According to claim 1, a synchronous positioning outdoor parking lot intelligent mapping method is characterized in that: The step of determining the driving trajectory of the vehicle according to the position information comprises: Based on a Kalman filter algorithm, Kalman position information is generated according to the position information; The Kalman position information is subjected to trajectory interpolation according to the Kalman position information to obtain a driving trajectory of the vehicle.

3. The method for intelligent mapping of outdoor parking lots with synchronous positioning according to claim 1, characterized in that: The step of determining the parking position of the vehicle according to the driving trajectory comprises: Acquire timestamp data corresponding to the driving trajectory according to the driving trajectory, and create an initial window according to a preset time length, the driving trajectory and the timestamp data; Sliding the initial window according to a preset time sliding unit to obtain at least one intermediate window; For each of the intermediate windows, determine whether the distances between all adjacent position points in the intermediate window are less than a preset distance based on the intermediate window, and determine whether multiple position points stay at the same position for more than a preset time. If so, determine the parking data of the vehicle based on the intermediate window, wherein the parking data includes the parking position, parking time period and parking duration.

4. The method for intelligent mapping of outdoor parking lots with synchronous positioning according to claim 1, characterized in that: The parking posture prediction model includes an input layer, a first feature extraction layer, a second feature extraction layer, a splicing layer, and an output layer. The step of generating the parking posture of the vehicle based on the parking posture prediction model according to the behavior monitoring data and the driving trajectory includes: Generate a first data feature vector according to the behavior monitoring data through the input layer; Generate a second data feature vector according to the driving trajectory through the input layer; generating a third data feature vector according to the first data feature vector through the first feature extraction layer; generating a fourth data feature vector according to the second data feature vector through the second feature extraction layer; The third data feature vector and the fourth data feature vector are concatenated through the concatenation layer to obtain a fifth data feature vector; Generate an output vector through the output layer according to the fifth data feature vector, wherein the output vector includes a first element, a second element, a third element and a fourth element, the first element is used to characterize the deflection angle between the front line of the vehicle and the first side of the reference parking space, the second element is used to characterize the deflection angle between the left side line of the vehicle and the second side of the reference parking space, the third element is used to characterize the deflection angle between the rear side line of the vehicle and the third side of the reference parking space, and the fourth element is used to characterize the deflection angle between the right side line of the vehicle and the fourth side of the reference parking space; A parking posture of the vehicle is generated based on the first element, the second element, the third element, and the fourth element.

5. The method for intelligent mapping of outdoor parking lots with synchronous positioning according to claim 4 is characterized in that: The first feature extraction layer includes a first fully connected layer, a second fully connected layer and a third fully connected layer, the ratio of the number of neurons in the first fully connected layer, the second fully connected layer and the third fully connected layer is 5:2:1, and the step of generating a third data feature vector according to the first data feature vector through the first feature extraction layer includes: Inputting the first data feature vector into the first fully connected layer to obtain a first intermediate feature vector; Inputting the first intermediate feature vector into the second fully connected layer to obtain a second intermediate feature vector; The second intermediate feature vector is input into the third fully connected layer to obtain a third data feature vector.

6. The method for intelligent mapping of outdoor parking lots with synchronous positioning according to claim 1, characterized in that: The step of determining the parking line data of the vehicle at the parking position according to the parking posture comprises: Acquire size information of the vehicle, wherein the size information includes length and width; Based on the parking space standard data of the parking lot and according to the size information, determining the parking space specification of the vehicle; Based on the parking posture of the vehicle and according to the parking space specification, parking line data of the vehicle at the parking position is determined.

7. A synchronous positioning outdoor parking lot intelligent mapping system, characterized in that: include: A data acquisition module, used to acquire the position information of the vehicle in real time through a positioning device after the vehicle enters the parking lot, and to acquire the behavior monitoring data of the vehicle through an on-board electronic system, wherein the behavior monitoring data includes at least one of on-board sensor data and driving operation data; a position information processing module, used to determine the driving trajectory of the vehicle according to the position information, and determine the parking position of the vehicle according to the driving trajectory; A parking posture generating module, configured to determine the parking posture of the vehicle based on a reference parking space and according to the behavior monitoring data and the driving trajectory, wherein the parking posture includes a parking angle; A parking space mapping map updating module, used to determine the parking line data of the vehicle at the parking position according to the parking posture, and update the parking space mapping map of the parking lot according to the parking line data; The step of determining the parking posture of the vehicle based on the reference parking space and according to the behavior monitoring data and the driving trajectory includes: Acquiring parking line data of the reference parking space based on the reference parking space, wherein the parking line data includes direction data of the parking line; Acquire model training data, and divide the model training data into a training set and a test set according to a preset ratio, wherein the model training data includes the parking line data, historical behavior detection data, and historical driving trajectory; The hyperparameters of the pre-built deep learning model were set according to the random network search algorithm, and the root mean square error RMSE and the coefficient of determination R 2 As an evaluation indicator; Training the deep learning model according to the training set to obtain a trained deep learning model; Testing the trained deep learning model according to the test set, and judging whether the error is within a preset range according to the evaluation index, and if so, using the trained deep learning model as a parking posture prediction model; Based on a parking posture prediction model, the parking posture of the vehicle is generated according to the behavior monitoring data and the driving trajectory.

8. The synchronous positioning outdoor parking lot intelligent mapping system according to claim 7 is characterized in that: The position information processing module includes: A Kalman filter submodule, for generating Kalman position information based on a Kalman filter algorithm and according to the position information; The trajectory interpolation submodule is used to perform trajectory interpolation on the Kalman position information according to the Kalman position information to obtain the driving trajectory of the vehicle.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor implements the method according to any one of claims 1 to 6 when executing the computer program.

Images