Vehicle positioning method in tunnel space and related device

Abstract

The invention relates to the technical field of vehicle positioning, and provides a vehicle positioning method in a tunnel space and a related device, and the method comprises the steps: obtaining a tunnel target point cloud data set; inputting the tunnel target point cloud data set into a Point RCNN network for processing to obtain preliminary tunnel target space three-dimensional positioning information; and according to the tunnel target point cloud data set and the preliminary tunnel target space three-dimensional positioning information, determining accurate tunnel target space three-dimensional positioning information by adopting an SLAM algorithm, and processing the tunnel target point cloud data set acquired in real time according to a Point RCNN network after the vehicle enters the tunnel. The accurate tunnel target space three-dimensional positioning information is determined in combination with the SLAM algorithm, and the positioning information of the target around the vehicle is obtained in real time, so that the real-time positioning information is provided for a vehicle driver, the driver can perform timely operation according to the positioning information, and the driving safety of the vehicle in the tunnel is improved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a vehicle positioning method and related device in a tunnel space. Background Art

[0002] In a modern transportation system, with the continuous increase in the number of vehicles, the issue of safe vehicle driving has been increasingly emphasized. Especially in the special scenario where a vehicle enters a tunnel, precise positioning of surrounding vehicles, people, and the environment and other targets is particularly important. Currently, after a vehicle enters a tunnel, in order to obtain positioning information about the surrounding environment, the Global Positioning System (GPS) technology is generally used. With its high precision and all-weather characteristics, the GPS technology can provide accurate vehicle positioning information in an open environment.

[0003] However, when a vehicle enters a closed or semi-closed environment such as a tunnel, due to the structural characteristics of the tunnel and signal laying cost limitations, GPS signals are often severely interfered with and attenuated. As a result, inside the tunnel, the reception of GPS signals becomes unstable, and signal loss may even occur, causing obvious delays and errors in the vehicle's positioning information about the surrounding environment. This lag and inaccuracy of the positioning information not only affect the normal operation of the vehicle navigation system but also may pose a potential threat to the driving safety of the vehicle. Summary of the Invention

[0004] The embodiments of this application provide a vehicle positioning method and related device in a tunnel space, which can process the real-time obtained tunnel target point cloud data set according to the PointRCNN network after the vehicle enters the tunnel, and determine the precise three-dimensional positioning information of the tunnel target space in combination with the SLAM algorithm, obtain the positioning information of the targets around the vehicle in real time, and improve the driving safety of the vehicle in the tunnel.

[0005] The first aspect of the embodiments of this application provides a vehicle positioning method in a tunnel space, and the method includes:

[0006] Obtain a tunnel target point cloud data set;

[0007] Input the tunnel target point cloud data set into the PointRCNN network for processing to obtain preliminary three-dimensional positioning information of the tunnel target space;

[0008] According to the tunnel target point cloud data set and the preliminary three-dimensional positioning information of the tunnel target space, use the SLAM algorithm to determine the precise three-dimensional positioning information of the tunnel target space.

[0009] In this example, first, the acquired tunnel target point cloud dataset is processed using the PointRCNN network to obtain preliminary three-dimensional positioning information of the tunnel target space. Then, based on the tunnel target point cloud dataset and the preliminary three-dimensional positioning information of the tunnel target space, the SLAM algorithm is further used for processing to obtain accurate three-dimensional positioning information of the tunnel target space. Therefore, the PointRCNN network and the SLAM algorithm can be used to process the real-time tunnel target point cloud dataset to obtain real-time and accurate positioning information of surrounding targets, providing a reference for the driver during driving in the tunnel, enabling the driver to make timely driving operations according to the positioning information, and avoiding the technical problem of possible delay that may occur when the traditional GPS technology enters the tunnel, which causes the positioning information obtained by the vehicle positioning function to be delayed simultaneously, resulting in a decrease in the accuracy of the positioning information, and improving the safety of the vehicle when driving in the tunnel.

[0010] Further, inputting the tunnel target point cloud dataset into the PointRCNN network for processing to obtain preliminary three-dimensional positioning information of the tunnel target space includes:

[0011] Preprocessing the tunnel target point cloud dataset to obtain a preprocessed tunnel target point cloud dataset;

[0012] Performing foreground cutting on the tunnel target point cloud dataset to obtain foreground point data and background point data;

[0013] Confirming the preliminary three-dimensional positioning information of the tunnel target space according to the foreground point data and the background point data.

[0014] Further, preprocessing the tunnel target point cloud dataset to obtain a preprocessed tunnel target point cloud dataset includes:

[0015] Using the downsampling method to extract key feature point data from the tunnel target point cloud dataset;

[0016] Performing data cleaning on the key feature point dataset to obtain key feature point data after data cleaning;

[0017] Extracting local region feature data from the key feature point dataset after data cleaning;

[0018] Performing data augmentation on the local region feature data to obtain enhanced local region feature data;

[0019] Recombining the enhanced local region feature data with the tunnel target point cloud dataset to obtain a preprocessed tunnel target point cloud dataset.

[0020] Further, perform data cleaning on the key feature point dataset to obtain the key feature point data after data cleaning, including:

[0021] Detect missing values and outliers in the key feature point dataset to obtain the detection results;

[0022] Process the missing values and outliers according to the detection results to obtain the complete key feature point data;

[0023] Perform standardization processing on the complete key feature point data using the Z-score method to obtain the key feature point data after data cleaning.

[0024] Further, according to the tunnel target point cloud dataset and the preliminary tunnel target space three-dimensional positioning information, use the SLAM algorithm to determine the accurate tunnel target space three-dimensional positioning information, including:

[0025] According to the tunnel target point cloud dataset, initialize the SLAM algorithm;

[0026] Extract feature point data from the preliminary tunnel target space three-dimensional positioning information to obtain a feature point dataset;

[0027] According to the feature point dataset, use the feature matching algorithm in the SLAM algorithm to match the current frame feature points and the previous frame feature points to obtain a feature point matching result;

[0028] According to the matching result, use the motion estimation algorithm in the SLAM algorithm to estimate the camera pose change;

[0029] Fuse the camera pose change with the preliminary tunnel target space three-dimensional positioning information to obtain the accurate tunnel target space three-dimensional positioning information.

[0030] The second aspect of the embodiments of the present application provides a vehicle tunnel space positioning device, and the device includes:

[0031] A first acquisition unit for acquiring a tunnel target point cloud dataset;

[0032] A first processing unit for inputting the tunnel target point cloud dataset into the PointRCNN network for processing to obtain preliminary tunnel target space three-dimensional positioning information;

[0033] A second processing unit for using the SLAM algorithm to determine the accurate tunnel target space three-dimensional positioning information according to the tunnel target point cloud dataset and the preliminary tunnel target space three-dimensional positioning information.

[0034] Further, in terms of determining the accurate three-dimensional positioning information of the tunnel target space by using the SLAM algorithm based on the tunnel target point cloud dataset and the preliminary three-dimensional positioning information of the tunnel target space, the second processing unit is configured to:

[0035] Based on the tunnel target point cloud dataset, initialize the SLAM algorithm;

[0036] Extract feature point data from the preliminary three-dimensional positioning information of the tunnel target space to obtain a feature point dataset;

[0037] Based on the feature point dataset, use the feature matching algorithm in the SLAM algorithm to match the current frame feature points and the previous frame feature points to obtain a feature point matching result;

[0038] Based on the matching result, use the motion estimation algorithm in the SLAM algorithm to estimate the camera pose change;

[0039] Fuse the camera pose change with the preliminary three-dimensional positioning information of the tunnel target space to obtain the accurate three-dimensional positioning information of the tunnel target space.

[0040] A third aspect of the embodiments of the present application provides a terminal, including a processor, an input device, an output device, and a memory. The processor, input device, output device, and memory are interconnected. Among them, the memory is used to store a computer program, and the computer program includes program instructions. The processor is configured to call the program instructions to execute the vehicle positioning method in the tunnel space as described in the first aspect.

[0041] A fourth aspect of the embodiments of the present application provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed by a processor, the processor is caused to execute the vehicle positioning method in the tunnel space as described in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0043] Figure 1 FIG. [ID] is a schematic diagram of the overall process of a vehicle positioning method in a tunnel space provided by an embodiment of the present application;

[0044] Figure 2This is a schematic flowchart of the preliminary three-dimensional positioning information of tunnel target space in a vehicle positioning method provided by an embodiment of the present application;

[0045] Figure 3 This is a schematic flowchart of the data preprocessing process of the tunnel target point cloud dataset in a vehicle positioning method provided by an embodiment of the present application;

[0046] Figure 4 This is a schematic flowchart of using the SLAM algorithm to process the preliminary three-dimensional positioning information of tunnel target space to obtain the accurate three-dimensional positioning information of tunnel target space in a vehicle positioning method provided by an embodiment of the present application;

[0047] Figure 5 This is a schematic structural diagram of a terminal provided by an embodiment of the present application;

[0048] Figure 6 This is a schematic structural diagram of a vehicle tunnel space positioning device provided by an embodiment of the present application. Detailed implementation manners

[0049] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0050] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.

[0051] Referring to "embodiment" in the present application means that a specific feature, structure or characteristic described in conjunction with the embodiment can be included in at least one embodiment of the present application. The phrase appears in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described in the present application can be combined with other embodiments.

[0052] To better understand a vehicle positioning method in a tunnel space provided by an embodiment of the present application, the scenarios where the vehicle positioning method in the tunnel space is applied will be briefly introduced below. When a vehicle drives into a tunnel, due to limited light conditions and restricted driver vision in the tunnel, the positioning function of the vehicle needs to be utilized to obtain the positioning information of vehicles or people around the vehicle. By referring to the positioning information, the driver can make driving operations in advance, thereby improving driving safety. Generally, the positioning function of an automobile relies on the Global Positioning System (GPS) technology. However, due to the structural characteristics of the tunnel and cost limitations of signal laying, GPS signals are often interfered with and attenuated. As a result, when the vehicle is driving inside the tunnel, the reception of GPS signals will have obvious delays and errors. Therefore, the positioning information obtained after the vehicle processes the GPS signals will have delays and errors, and there is a time delay in the positioning function of the automobile. The driver cannot timely perform corresponding driving operations by referring to the positioning information, reducing the driving safety of the vehicle in the tunnel. The embodiment of the present application aims to solve the problem that the driving safety of the vehicle is reduced due to poor GPS signals when the vehicle drives into the tunnel, and provides a vehicle positioning method in the tunnel space. This method can generate timely and accurate three-dimensional positioning information of the tunnel target space by obtaining tunnel target point cloud data and combining the PointRCNN network and the SLAM algorithm, improving the driving safety of the vehicle in the tunnel.

[0053] The vehicle positioning method in the tunnel space is applied to a vehicle tunnel space positioning device. Figure 1 Figure 1 shows a schematic diagram of the overall process of a vehicle positioning method in a tunnel space. As Figure 1 shown, it specifically includes:

[0054] S1. Obtain a tunnel target point cloud data set.

[0055] Specifically, the LiDAR sensor arranged on the vehicle is used to collect the point cloud data of the real-time surrounding environment, including the point cloud data of key targets such as vehicles and people, providing a data source for subsequent data processing.

[0056] S2. Input the tunnel target point cloud data set into the PointRCNN network for processing to obtain preliminary three-dimensional positioning information of the tunnel target space.

[0057] Specifically, as Figure 2 shown, a schematic diagram of the process of preliminary three-dimensional positioning information of the tunnel target space in a vehicle positioning method in a tunnel space is provided, including:

[0058] S2-1. Preprocess the tunnel target point cloud data set to obtain a preprocessed tunnel target point cloud data set. As a preferred solution of this embodiment, as Figure 3As shown, a schematic diagram of the data preprocessing process for the tunnel target point cloud dataset in a vehicle positioning method within a tunnel space is provided, including:

[0059] S2-1-1. Extract key feature point data from the tunnel target point cloud dataset using downsampling. Since the original point cloud data is often very dense and large in quantity, directly processing this data would consume a large amount of computing resources. Therefore, downsampling is first performed to retain the key feature points, thereby reducing the computational amount and accelerating the subsequent processing process.

[0060] S2-1-2. Clean the key feature point dataset to obtain the key feature point data after data cleaning.

[0061] Specifically, the data cleaning steps include:

[0062] Detect missing values and outliers in the key feature point dataset to obtain a detection result, which indicates the missing values and outliers in the current key feature point dataset.

[0063] According to the detection result, process the missing values and outliers to obtain complete key feature point data. After processing, perform row denoising on the complete key feature point data to remove outliers and noise points to ensure data quality.

[0064] Standardize the complete key feature point data using the Z-score method to obtain the key feature point data after data cleaning.

[0065] S2-1-3. Extract local region feature data from the key feature point dataset after data cleaning. Since the key feature screening of the point cloud data is performed in the process of this embodiment, the sample quantity is small, so it is necessary to expand the key feature point data.

[0066] S2-1-4. Augment the local region feature data to obtain the augmented local region feature data.

[0067] S2-1-5. Recombine the augmented local region feature data with the tunnel target point cloud dataset to obtain the preprocessed tunnel target point cloud dataset. Therefore, the integrity and quantity of the preprocessed tunnel target point cloud dataset meet the requirements of subsequent processing.

[0068] S2-2. Perform foreground cutting on the tunnel target point cloud dataset to obtain foreground point data and background point data. The foreground points correspond to target objects such as vehicles and pedestrians in the tunnel, while the background points correspond to static scenes such as tunnel walls and the ground.

[0069] S2-3. Confirm the preliminary three-dimensional positioning information of the tunnel target space according to the foreground point data and the background point data.

[0070] Specifically, this embodiment preferably includes:

[0071] A1. Generate a 3D candidate region according to the foreground point data and the background point data.

[0072] Based on the extracted features and the foreground segmentation results, the network generates multiple 3D candidate boxes. (5) Each candidate box represents the possible position and size of the target. According to the feature information in the point cloud, the possible position area of the target object is speculated, and the corresponding 3D candidate box is generated. The regression loss based on Bin is introduced to constrain the generation of the candidate box. The regression branch only regresses the 3D bounding box position of the foreground points during training. Finally, the IOU of the 3D box of each foreground point is calculated to filter out the candidate boxes with a large overlap. The specific calculation is as follows:

[0073]

[0074] Among them, S represents the search range set to 3; δ represents the interval length of 0.5. In order to accurately locate the target center position from the discrete point cloud, the search range S on the X and Z axes is evenly divided into intervals of δ, representing the center positions of each target to be measured on the X-Z plane; x (p) , y (p) , z (p) are the foreground point coordinates of the target to be measured; x p , y p , z p are the center coordinates of its corresponding target; and are the true box intervals allocated along the X axis and the Z axis; is the true box residual for further positioning within the allocated partition; c is the normalized partition length.

[0075] A2. Perform point cloud region pooling operation according to multiple 3D candidate regions.

[0076] To more accurately capture the geometric and semantic information of the target object, for each 3D candidate box, a point cloud region pooling operation is performed to extract local features from the point cloud within the candidate box. Expand the range of the candidate box by a constant n, that is, the box. The example operation is as follows:

[0077] Let b i =(x i , y i , z i , h i , w i , l i , θ i ) become

[0078] For aggregate the features of all points inside, including: the original point cloud coordinates (x, y, z), the reflection intensity (r), the category of this point, and the feature vector of the C dimension.

[0079] A3. Perform coordinate transformation.

[0080] To further improve the detection accuracy, the network will transform the point cloud within the candidate box into the canonical space. The canonical space is a standardized coordinate system, which helps the network learn more robust local features.

[0081] A4. Optimize the 3D candidate region bounding box.

[0082] The initially generated 3D candidate bounding boxes are usually rough. To improve the localization accuracy, it is necessary to optimize the candidate bounding boxes. Combining the global semantic features and local spatial features, refine the candidate bounding boxes, including adjusting parameters such as position, size, and orientation.

[0083] A5. Post-process the optimized 3D candidate region.

[0084] After generating the final detection result bounding box, the network will apply non-maximum suppression (NMS) to remove redundant detection results. NMS compares the overlap degree (IoU) between detection bounding boxes, retains the detection bounding box with the highest confidence, and removes the detection bounding boxes with a higher overlap degree.

[0085] A6. Set the bounding box optimization loss function for the 3D candidate region.

[0086] In the embodiment of this application, if the IOU between the selected bounding box and the ground truth bounding box exceeds 0.55, the candidate bounding box is used as the learning target. The loss function is as follows: the first term is the classification loss of positive and negative samples; the second term is the position refinement loss of positive samples.

[0087]

[0088] Among them, B represents the total number of samples in a batch; B pos represents the total number of positive samples for regression in a batch; prob i is the confidence of the predicted positive sample, and label i is the corresponding label information; F cls is the cross-entropy loss used to supervise the predicted confidence; and correspond to the position regression and size regression after refinement of the 3D candidate bounding box generated in the first stage respectively.

[0089] A7. Perform object classification and regression in the 3D candidate region.

[0090] Specifically include:

[0091] First, based on the optimized 3D candidate boxes, the point cloud data within each candidate box is further processed to identify specific target categories. For example, the system can identify that the targets within the candidate boxes are vehicles, pedestrians, devices, etc.

[0092] Then, 3D position regression is performed. For each target, 3D position regression is carried out to calculate the precise spatial coordinates, dimensions (length, width, height), and orientation angle of the target, and to determine its specific position and pose in the tunnel space.

[0093] Furthermore, the three-dimensional coordinate information is output, and finally the 3D position information of each detected target is output, including data such as the center point coordinates, dimensions (length, width, height), and orientation angle of the target, obtaining the preliminary 3D positioning information of the tunnel target space.

[0094] S3. According to the tunnel target point cloud dataset and the preliminary 3D positioning information of the tunnel target space, use the SLAM algorithm to determine the precise 3D positioning information of the tunnel target space.

[0095] Specifically, as Figure 4 shown, a schematic flow chart of using the SLAM algorithm to process the preliminary 3D positioning information of the tunnel target space to obtain the precise 3D positioning information of the tunnel target space in a vehicle positioning method within the tunnel space is provided, including:

[0096] S3-1. According to the tunnel target point cloud dataset, initialize the SLAM algorithm.

[0097] Specifically, first, obtain the tunnel target point cloud dataset from a tunnel scanning device such as a lidar. This dataset contains a large amount of three-dimensional point information about the internal space of the tunnel. Then, select a suitable SLAM (Simultaneous Localization and Mapping) algorithm framework for initialization, including setting the number of iterations and optimization threshold of the algorithm, and loading the filtering module and denoising module to ensure the accuracy and efficiency of subsequent processing. After initialization, the algorithm enters the iterative optimization stage and is ready to process the input point cloud data.

[0098] S3-2. Extract feature point data from the preliminary 3D positioning information of the tunnel target space to obtain a feature point dataset.

[0099] Specifically, in the preliminary 3D positioning information of the tunnel target space, the ORB feature extraction algorithm is used to identify points with significant local features from the point cloud data. These points are usually key points that can be stably identified from different perspectives. During the feature extraction process, the descriptor of each feature point, which is a vector, is calculated to describe the geometric or texture information around the point. These feature points and their descriptors are combined to form a feature point dataset, preparing for subsequent feature matching.

[0100] S3-3. According to the feature point dataset, use the feature matching algorithm in the SLAM algorithm to match the feature points of the current frame and the previous frame to obtain the feature point matching result.

[0101] Specifically, use the FLANN matching feature matching algorithm to search for the most similar feature points in the feature point dataset of the current frame (i.e., the latest acquired frame of point cloud data) in the feature point dataset of the previous frame. The matching process is based on the similarity of the feature point descriptors and is achieved by calculating the Euclidean distance or Hamming distance between the descriptors. After the matching is completed, a series of feature point pairs are obtained, which represent the feature positions that can be stably tracked between consecutive frames and provide key information for subsequent motion estimation.

[0102] S3-4. According to the matching result, use the motion estimation algorithm in the SLAM algorithm to estimate the camera pose change.

[0103] Specifically, based on the feature point matching result, use the ICP algorithm motion estimation algorithm to estimate the camera pose change. The pose change includes the rotation matrix and translation vector of the camera, which describe the motion of the camera from the previous frame to the current frame. The algorithm iteratively optimizes the pose estimation by minimizing the reprojection error between the feature points until the preset convergence condition or the upper limit of the number of iterations is reached. The optimized pose change reflects the actual motion trajectory of the camera in the tunnel space.

[0104] S3-5. Fuse the camera pose change with the preliminary 3D positioning information of the tunnel target space to obtain the accurate 3D positioning information of the tunnel target space.

[0105] Specifically, combine the estimated camera pose change with the preliminary 3D positioning information of the tunnel target space, and integrate the new position information into the global map through coordinate transformation and map update strategies. This usually involves transforming the point cloud data of the current frame according to the camera pose change and then fusing it with the global map to update and optimize the 3D model of the entire tunnel space. During the fusion process, Gaussian filtering is applied to smooth the noise, and the BA map optimization algorithm is used to further reduce the cumulative error. Finally, the accurate 3D positioning information of the tunnel target space is obtained.

[0106] That is, in this embodiment, the obtained tunnel target point cloud dataset is processed by the PointRCNN network to obtain preliminary three-dimensional positioning information of the tunnel target space. Then, according to the tunnel target point cloud dataset and the preliminary three-dimensional positioning information of the tunnel target space, the SLAM algorithm is used for further processing to obtain accurate three-dimensional positioning information of the tunnel target space. Therefore, the PointRCNN network and the SLAM algorithm can be used to process the real-time tunnel target point cloud dataset to obtain real-time and accurate positioning information of surrounding targets, providing a reference for the driver during driving in the tunnel, enabling the driver to make timely driving operations based on the positioning information, and avoiding the problem that the positioning information obtained by the vehicle positioning function may be delayed due to the possible delay when the GPS signal enters the tunnel, resulting in a decrease in the accuracy of the positioning information, thus improving the safety of the vehicle when driving in the tunnel.

[0107] Consistent with the above embodiment, please refer to Figure 5 , Figure 5 which is a schematic structural diagram of a terminal provided by an embodiment of the present application. As shown in the figure, it includes a processor, an input device, an output device, and a memory. The processor, the input device, the output device, and the memory are interconnected. Among them, the memory is used to store a computer program, and the computer program includes program instructions. The processor is configured to call the program instructions, and the above program includes instructions for performing the following steps;

[0108] Obtain a tunnel target point cloud dataset;

[0109] Input the tunnel target point cloud dataset into the PointRCNN network for processing to obtain preliminary three-dimensional positioning information of the tunnel target space;

[0110] According to the tunnel target point cloud dataset and the preliminary three-dimensional positioning information of the tunnel target space, use the SLAM algorithm to determine accurate three-dimensional positioning information of the tunnel target space.

[0111] In this example, the obtained tunnel target point cloud dataset is processed using the PointRCNN network to obtain preliminary three-dimensional positioning information of the tunnel target space. Then, based on the tunnel target point cloud dataset and the preliminary three-dimensional positioning information of the tunnel target space, the SLAM algorithm is used for further processing to obtain accurate three-dimensional positioning information of the tunnel target space. Therefore, the PointRCNN network and the SLAM algorithm can be used to process the real-time tunnel target point cloud dataset to obtain real-time and accurate positioning information of surrounding targets, providing a reference for the driver during the driving process in the tunnel, enabling the driver to make timely driving operations based on the positioning information, avoiding the technical problem of possible delays that may occur when the traditional GPS technology enters the tunnel, which causes the positioning information obtained by the vehicle positioning function to be delayed simultaneously, resulting in a decrease in the accuracy of the positioning information, and improving the safety of the vehicle when driving in the tunnel.

[0112] The above mainly introduces the solution of the embodiment of the present application from the perspective of the execution process on the method side. It can be understood that in order for the terminal to implement the above functions, it includes the corresponding hardware structure and / or software module for executing each function. Those skilled in the art should easily realize that, combining the units and algorithm steps of each example described in the embodiments provided in this article, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the way of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0113] The embodiment of the present application can divide the functions of the terminal according to the above method examples. For example, each function unit can be divided corresponding to each function, or two or more functions can be integrated into one processing unit. The above integrated unit can be implemented in the form of hardware or in the form of a software function unit. It should be noted that the division of units in the embodiment of the present application is illustrative, only a logical function division, and there may be other division methods in actual implementation.

[0114] Consistent with the above, please refer to Figure 6 , Figure 6 which is a schematic structural diagram of a service product determination device provided by an embodiment of the present application. As Figure 6 shown, the device includes:

[0115] The first acquisition unit 1 is used to acquire a tunnel target point cloud dataset;

[0116] The first processing unit 2 is configured to input the tunnel target point cloud dataset into the PointRCNN network for processing to obtain preliminary three-dimensional positioning information of the tunnel target space;

[0117] The second processing unit 3 is configured to determine the accurate three-dimensional positioning information of the tunnel target space by using the SLAM algorithm according to the tunnel target point cloud dataset and the preliminary three-dimensional positioning information of the tunnel target space.

[0118] In a possible implementation manner, in the aspect of inputting the tunnel target point cloud dataset into the PointRCNN network for processing to obtain preliminary three-dimensional positioning information of the tunnel target space, the first processing unit is configured to:

[0119] Preprocess the tunnel target point cloud dataset to obtain a preprocessed tunnel target point cloud dataset;

[0120] Perform foreground cutting on the tunnel target point cloud dataset to obtain foreground point data and background point data;

[0121] Confirm the preliminary three-dimensional positioning information of the tunnel target space according to the foreground point data and the background point data.

[0122] In a possible implementation manner, the device is further configured to:

[0123] Extract key feature point data from the tunnel target point cloud dataset by using the downsampling method;

[0124] Perform data cleaning on the key feature point dataset to obtain key feature point data after data cleaning;

[0125] Extract local region feature data from the key feature point dataset after data cleaning;

[0126] Perform data augmentation on the local region feature data to obtain enhanced local region feature data;

[0127] Recombine the enhanced local region feature data with the tunnel target point cloud dataset to obtain a preprocessed tunnel target point cloud dataset.

[0128] In a possible implementation manner, the device is further configured to:

[0129] Detect missing values and outliers in the key feature point dataset to obtain a detection result;

[0130] Process the missing values and outliers according to the detection result to obtain complete key feature point data;

[0131] The Z-score method is used to standardize the complete key feature point data, and the key feature point data after data cleaning is obtained.

[0132] In the aspect of determining the accurate three-dimensional positioning information of the tunnel target space by using the SLAM algorithm according to the tunnel target point cloud data set and the preliminary three-dimensional positioning information of the tunnel target space, the second processing unit is used for:

[0133] According to the tunnel target point cloud data set, initialize the SLAM algorithm;

[0134] Extract the feature point data from the preliminary three-dimensional positioning information of the tunnel target space to obtain a feature point data set;

[0135] According to the feature point data set, use the feature matching algorithm in the SLAM algorithm to match the current frame feature points and the previous frame feature points to obtain a feature point matching result;

[0136] According to the matching result, use the motion estimation algorithm in the SLAM algorithm to estimate the change of the camera pose;

[0137] Fuse the change of the camera pose with the preliminary three-dimensional positioning information of the tunnel target space to obtain the accurate three-dimensional positioning information of the tunnel target space.

[0138] The embodiment of the present application also provides a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, and the computer program enables the computer to execute some or all of the steps of any one of the vehicle positioning methods in the tunnel space described in the above method embodiments.

[0139] The embodiment of the present application also provides a computer program product, and the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program enables the computer to execute some or all of the steps of any one of the vehicle positioning methods in the tunnel space described in the above method embodiments.

[0140] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0141] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0142] In several embodiments provided by the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling, direct coupling, or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical or other form.

[0143] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0144] In addition, each functional unit in the various embodiments of the application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software program module.

[0145] If the above-mentioned integrated unit is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. And the aforementioned memory includes: USB flash drive, read-only memory (ROM), random access memory (RAM), mobile hard disk, magnetic disk, or optical disk, etc., which can store program codes.

[0146] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program. This program can be stored in a computer-readable memory. The memory can include: flash drive, read-only memory, random access memory, magnetic disk, or optical disk, etc.

[0147] The above has introduced the embodiments of the present application in detail. Specific examples are used herein to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A vehicle positioning method in a tunnel space, characterized in that, Including: Obtain the tunnel target point cloud dataset; Input the tunnel target point cloud dataset into the PointRCNN network for processing to obtain preliminary three-dimensional spatial positioning information of the tunnel target; According to the tunnel target point cloud dataset and the preliminary three-dimensional spatial positioning information of the tunnel target, use the SLAM algorithm to determine the accurate three-dimensional spatial positioning information of the tunnel target.

2. The vehicle positioning method in a tunnel space according to claim 1, wherein Input the tunnel target point cloud dataset into the PointRCNN network for processing to obtain preliminary three-dimensional spatial positioning information of the tunnel target, including: Preprocess the tunnel target point cloud dataset to obtain a preprocessed tunnel target point cloud dataset; Perform foreground cutting on the tunnel target point cloud dataset to obtain foreground point data and background point data; Confirm the preliminary three-dimensional spatial positioning information of the tunnel target according to the foreground point data and the background point data.

3. The vehicle positioning method in the tunnel space according to claim 2, wherein Preprocess the tunnel target point cloud dataset to obtain a preprocessed tunnel target point cloud dataset, including: Use the downsampling method to extract key feature point data from the tunnel target point cloud dataset; Clean the key feature point dataset to obtain key feature point data after data cleaning; Extract local region feature data from the key feature point dataset after data cleaning; Perform data augmentation on the local region feature data to obtain enhanced local region feature data; Recombine the enhanced local region feature data with the tunnel target point cloud dataset to obtain a preprocessed tunnel target point cloud dataset.

4. The vehicle positioning method in the tunnel space according to claim 3, characterized in that, Clean the key feature point dataset to obtain key feature point data after data cleaning, including: Detect missing values and outliers in the key feature point dataset to obtain a detection result; Process the missing values and outliers according to the detection result to obtain complete key feature point data; Use the Z-score method to standardize the complete key feature point data to obtain key feature point data after data cleaning.

5. The vehicle positioning method in the tunnel space according to claim 1, characterized in that, According to the tunnel target point cloud dataset and the preliminary three-dimensional spatial positioning information of the tunnel target, use the SLAM algorithm to determine the accurate three-dimensional spatial positioning information of the tunnel target, including: Initialize the SLAM algorithm according to the tunnel target point cloud dataset; Extract feature point data from the preliminary three-dimensional spatial positioning information of the tunnel target to obtain a feature point dataset; According to the feature point dataset, use the feature matching algorithm in the SLAM algorithm to match the current frame feature points and the previous frame feature points to obtain a feature point matching result; According to the matching result, use the motion estimation algorithm in the SLAM algorithm to estimate the camera pose change; Fuse the camera pose change with the preliminary three-dimensional spatial positioning information of the tunnel target to obtain the accurate three-dimensional spatial positioning information of the tunnel target.

6. A vehicle tunnel space positioning device, characterized in that, The device includes: A first acquisition unit for acquiring the tunnel target point cloud dataset; A first processing unit for inputting the tunnel target point cloud dataset into the PointRCNN network for processing to obtain preliminary three-dimensional spatial positioning information of the tunnel target; A second processing unit, configured to determine accurate three-dimensional positioning information of the tunnel target space by using a SLAM algorithm according to the tunnel target point cloud data set and the preliminary three-dimensional positioning information of the tunnel target space.

7. The vehicle tunnel space positioning device according to claim 6, wherein, Regarding inputting the tunnel target point cloud data set into the PointRCNN network for processing to obtain preliminary three-dimensional positioning information of the tunnel target space, the first processing unit is configured to: Preprocess the tunnel target point cloud data set to obtain a preprocessed tunnel target point cloud data set; Perform foreground segmentation on the tunnel target point cloud data set to obtain foreground point data and background point data; Confirm the preliminary three-dimensional positioning information of the tunnel target space according to the foreground point data and the background point data.

8. The vehicle tunnel space positioning device according to claim 7, characterized in that, Regarding determining accurate three-dimensional positioning information of the tunnel target space by using a SLAM algorithm according to the tunnel target point cloud data set and the preliminary three-dimensional positioning information of the tunnel target space, the second processing unit is configured to: Initialize the SLAM algorithm according to the tunnel target point cloud data set; Extract feature point data from the preliminary three-dimensional positioning information of the tunnel target space to obtain a feature point data set; According to the feature point data set, use the feature matching algorithm in the SLAM algorithm to match the current frame feature points and the previous frame feature points to obtain a feature point matching result; According to the matching result, use the motion estimation algorithm in the SLAM algorithm to estimate the camera pose change; Fuse the camera pose change with the preliminary three-dimensional positioning information of the tunnel target space to obtain accurate three-dimensional positioning information of the tunnel target space.

9. A terminal, characterized in that, It includes a processor, an input device, an output device, and a memory. The processor, input device, output device, and memory are interconnected. Among them, the memory is used to store a computer program, and the computer program includes program instructions. The processor is configured to call the program instructions to execute the vehicle positioning method in the tunnel space according to any one of claims 1-5.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed by the processor, the processor is caused to execute the vehicle positioning method in the tunnel space according to any one of claims 1-5.

Images