A method and apparatus for Bayesian trajectory confidence estimation based on crowdsourced map aggregation

Abstract

This invention relates to a Bayesian trajectory confidence estimation method and apparatus based on crowdsourced map aggregation, comprising: registering and tile-slicing map fragments using the ICP algorithm based on trajectory data sampled multiple times; rasterizing the tiled map fragments and generating rasterized trajectory density according to the trajectory direction; forming road vectors based on the raster and trajectory direction and establishing a correlation with the original trajectory data; fitting a probability model based on the density generated when generating road vectors and sampling to obtain a posterior distribution; and calculating the trajectory confidence value based on the posterior distribution. This invention significantly improves computational efficiency by using trajectory density and direction division as the primary input format, rather than directly substituting trajectory coordinates and other information into a complex model for calculation.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, and specifically to a method and apparatus for estimating Bayesian trajectory confidence values ​​based on crowdsourced map aggregation. Background Technology

[0002] Current confidence-based techniques for high-precision map trajectories have not yet reached industry standards. They primarily rely on iterative map-trajectory co-optimization processes, aiming to simultaneously improve the quality of map updates and map matching. This framework is based on map matching and updates, and incorporates concepts such as confidence values, map influence, and matching quality into the co-optimization process to optimize deep neural network computation. These methods require significant computational resources and have strict requirements for data reduction, hindering the production and updating of large numbers of high-precision maps. Other approaches use high-precision trajectory data-driven methods, combining traffic conditions and path conflicts, and using planning control to mine the internal relationships within the high-precision trajectory data to assess lane-level real-time traffic safety. However, such research is scarce, and these methods also have strict requirements for datasets and are difficult to generalize. Summary of the Invention

[0003] This invention addresses the technical problems existing in the prior art by providing a Bayesian trajectory confidence value estimation method and apparatus based on crowdsourced map aggregation, focusing on how to simply and efficiently estimate the confidence value of a large amount of crowdsourced trajectory data using fewer features.

[0004] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0005] In a first aspect, the present invention provides a Bayesian trajectory confidence value estimation method based on crowdsourced map aggregation, comprising:

[0006] Based on trajectory data sampled multiple times, the ICP algorithm is used to register map fragments and cut them into tiles;

[0007] The tiled map fragments are rasterized and rasterized trajectory density is generated according to the trajectory direction;

[0008] A road vector is generated based on the grid and trajectory direction, and a correlation is established with the original trajectory data.

[0009] The posterior distribution is obtained by fitting a probability model based on the density of the generated road vectors and sampling. The confidence value of the trajectory is then calculated based on the posterior distribution.

[0010] Furthermore, the process of registering and tile-slicing map fragments using the ICP algorithm based on multiple sampled trajectory data includes:

[0011] Initialize the rotation matrix as the identity matrix R0 = I, and find the set of closest points between any two sets of trajectory data using the KD-tree spatial index;

[0012] Based on the set of nearest points, the least squares method is used to solve and update the rotation matrix R*; using the rotation matrix R*, either of the two sets of trajectory data is rotated, and the set of nearest points in the two sets of trajectory data is found again according to the KD-tree spatial index; this step is repeated until the rotation and translation amount of the rotation matrix R* is less than a preset threshold, then the iteration stops and the map fragment registration is completed.

[0013] The registered map fragments are tiled according to a specified ratio.

[0014] Furthermore, the step of rasterizing the tiled map fragments and generating rasterized trajectory density based on the trajectory direction includes:

[0015] The tiled map fragments are rasterized, with each raster representing a pixel.

[0016] The rasterized map fragments are divided into four layers according to the trajectory direction. The four layers correspond to southwest to northeast, northeast to southwest, northwest to southeast, and southeast to northwest, respectively. Each layer contains only trajectory data whose trajectory direction matches the corresponding direction of the layer.

[0017] In each layer, the trajectory pixels contained in that direction are superimposed one by one to form a rasterized trajectory density.

[0018] Furthermore, the method of rasterizing the tiled map fragments and generating rasterized trajectory density according to the trajectory direction also includes:

[0019] For each layer, a convolutional layer of preset size is used to smooth the rasterized map fragments.

[0020] Furthermore, the step of forming a road vector based on the grid and trajectory direction and establishing a correlation with the original trajectory data includes:

[0021] For any given layer, select the pixel P with the highest trajectory density in that layer;

[0022] Starting from pixel P, find the pixel P' with the smallest trajectory density difference from the starting point along the corresponding direction of the layer; then, along the direction of P→P', find the pixel with the smallest trajectory density difference from the previous pixel pixel pixel by pixel until the edge of the map fragment, the pixel with a density of 0, or the pixel with a density less than the preset threshold B, and generate the first set.

[0023] Starting from pixel P, find the pixel with the smallest trajectory density difference from the starting point along the opposite direction to the layer direction; then, along the direction of P→P”, find the pixel with the smallest trajectory density difference from the previous pixel pixel one by one until the edge of the map fragment, the pixel with a density of 0, or the pixel with a density less than the preset threshold B, and generate the second set.

[0024] Connect the pixels in the first and second sets along the corresponding direction of the layer to generate a road vector;

[0025] The KD-tree spatial indexing method is used to find the trajectory associated with the road vector and establish a relationship with its trajectory ID.

[0026] Furthermore, a probability model is fitted based on the density during road vector generation, and a posterior distribution is obtained by sampling. The confidence value of the trajectory is then calculated based on the posterior distribution, including:

[0027] Based on the relationship between road vectors and trajectories, the number of trajectories associated with each road vector is counted. The number of trajectories is fitted with a Poisson distribution as the likelihood, and then a gamma distribution is used as the conjugate prior distribution. The hyperparameters of the gamma distribution are sampled from the uninformed prior distribution or estimated from the data. Then, the posterior distribution is sampled according to Markov Monte Carlo or other Bayesian methods, and the empirical cumulative distribution is obtained to obtain its confidence value.

[0028] Secondly, the present invention provides a Bayesian trajectory confidence value estimation device based on crowdsourced map aggregation, comprising:

[0029] The trajectory location registration module uses the ICP algorithm to register and tile map fragments based on multiple sampled trajectory data.

[0030] The layered rasterization module rasterizes the tiled map fragments and generates rasterized trajectory density based on the trajectory direction.

[0031] The vector trajectory association module generates road vectors based on the grid and trajectory direction and establishes a relationship with the original trajectory data;

[0032] The confidence value calculation module fits a probability model based on the density when generating road vectors and obtains the posterior distribution by sampling. It then calculates the confidence value of the trajectory based on the posterior distribution.

[0033] Thirdly, the present invention provides an electronic device, comprising:

[0034] Memory, used to store computer software programs;

[0035] A processor is configured to read and execute the computer software program, thereby implementing the Bayesian trajectory confidence estimation method based on crowdsourced map aggregation as described in the first aspect of the present invention.

[0036] Fourthly, the present invention provides a non-transitory computer-readable storage medium storing a computer software program, which, when executed by a processor, implements a Bayesian trajectory confidence estimation method based on crowdsourced map aggregation as described in the first aspect of the present invention.

[0037] The beneficial effects of this invention are: by using trajectory density and direction as the primary input format and employing convolution for smoothing, instead of directly substituting trajectory coordinates and other information into a complex model for calculation, this invention can significantly improve computational efficiency. The data required by this invention can be obtained from various trajectory GPS acquisition devices, and apart from basic direction and position, there are no other special requirements, demonstrating strong generalization ability. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of a Bayesian trajectory confidence value estimation method based on crowdsourced map aggregation, provided in an embodiment of the present invention.

[0039] Figure 2 This is a schematic diagram of a Bayesian trajectory confidence value estimation device based on crowdsourced map aggregation, provided in an embodiment of the present invention.

[0040] Figure 3 A schematic diagram of an embodiment of the electronic device provided in this invention;

[0041] Figure 4 This is a schematic diagram of an embodiment of a computer-readable storage medium provided in this invention. Detailed Implementation

[0042] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0043] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0044] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0045] like Figure 1 As shown, this embodiment of the invention provides a Bayesian trajectory confidence estimation method based on crowdsourced map aggregation, including:

[0046] S100 uses the ICP algorithm to register and tile map fragments based on multiple sampled trajectory data.

[0047] The trajectories in map fragments obtained from multiple samplings will change, therefore trajectory registration is necessary. Specifically, the ICP algorithm is used for trajectory registration. First, the rotation matrix is ​​initialized as the identity matrix R0 = I. The nearest point set between any two sets of trajectory data is found using the KD-tree spatial index. Then, based on this nearest point set, the rotation matrix R* is updated using the least squares method. Using the rotation matrix R*, either set of trajectory data is rotated, and the nearest point set between the two sets of trajectory data is found again using the KD-tree spatial index. This step is repeated iteratively until the rotation and translation amount of the rotation matrix R* is less than a preset threshold, at which point the iteration stops, and map fragment registration is complete. Finally, the registered map fragments are tiled according to a specified ratio, for example, divided into 500m * 500m areas.

[0048] S200 rasterizes the tiled map fragments and generates rasterized trajectory density based on the trajectory direction.

[0049] The tiled map fragments are rasterized, with each raster representing a pixel.

[0050] The rasterized map fragments are divided into four layers according to the trajectory direction. The four layers correspond to southwest to northeast, northeast to southwest, northwest to southeast, and southeast to northwest, respectively. Each layer contains only trajectory data whose trajectory direction matches the corresponding direction of the layer.

[0051] In each layer, the trajectory pixels contained in that direction are superimposed one by one to form a rasterized trajectory density.

[0052] For each layer, a convolutional layer of preset size is used to smooth the rasterized map fragments.

[0053] S300 generates a road vector based on the grid and trajectory direction and establishes a correlation with the original trajectory data.

[0054] For any given layer, select the pixel P with the highest trajectory density in that layer;

[0055] Starting from pixel P, find the pixel P' with the smallest trajectory density difference from the starting point along the corresponding direction of the layer; then, along the direction of P→P', find the pixel with the smallest trajectory density difference from the previous pixel pixel pixel by pixel until the edge of the map fragment, the pixel with a density of 0, or the pixel with a density less than the preset threshold B, and generate the first set.

[0056] Starting from pixel P, find the pixel with the smallest trajectory density difference from the starting point along the opposite direction to the layer direction; then, along the direction of P→P”, find the pixel with the smallest trajectory density difference from the previous pixel pixel one by one until the edge of the map fragment, the pixel with a density of 0, or the pixel with a density less than the preset threshold B, and generate the second set.

[0057] Connect the pixels in the first and second sets along the corresponding direction of the layer to generate a road vector;

[0058] The KD-tree spatial indexing method is used to find the trajectory associated with the road vector and establish a relationship with its trajectory ID.

[0059] S400: Fit a probability model based on the density when generating road vectors and obtain the posterior distribution by sampling; calculate the confidence value of the trajectory based on the posterior distribution.

[0060] Based on the relationship between road vectors and trajectories, the number of trajectories associated with each road vector is counted. The number of trajectories is fitted with a Poisson distribution as the likelihood, and then a gamma distribution is used as the conjugate prior distribution. The hyperparameters of the gamma distribution are sampled from the uninformed prior distribution or estimated from the data. Then, the posterior distribution is sampled according to Markov Monte Carlo or other Bayesian methods, and the empirical cumulative distribution is obtained to obtain its confidence value.

[0061] Secondly, the present invention provides a Bayesian trajectory confidence value estimation device based on crowdsourced map aggregation, comprising:

[0062] The trajectory location registration module uses the ICP algorithm to register and tile map fragments based on multiple sampled trajectory data.

[0063] The layered rasterization module rasterizes the tiled map fragments and generates rasterized trajectory density based on the trajectory direction.

[0064] The vector trajectory association module generates road vectors based on the grid and trajectory direction and establishes a relationship with the original trajectory data;

[0065] The confidence value calculation module fits a probability model based on the density when generating road vectors and obtains the posterior distribution by sampling. It then calculates the confidence value of the trajectory based on the posterior distribution.

[0066] Please see Figure 3 , Figure 3 This is a schematic diagram illustrating an embodiment of the electronic device provided in this invention. For example... Figure 3 As shown, an embodiment of the present invention provides an electronic device 500, including a memory 510, a processor 520, and a computer program 511 stored in the memory 510 and executable on the processor 520. When the processor 520 executes the computer program 511, it performs the following steps:

[0067] S100: Based on the trajectory data sampled multiple times, the ICP algorithm is used to register map fragments and cut them into tiles.

[0068] S200 rasterizes the tiled map fragments and generates rasterized trajectory density based on the trajectory direction;

[0069] S300 generates a road vector based on the grid and trajectory direction and establishes a correlation with the original trajectory data;

[0070] S400: Fit a probability model based on the density when generating road vectors and obtain the posterior distribution by sampling; calculate the confidence value of the trajectory based on the posterior distribution.

[0071] Please see Figure 4 , Figure 4 This is a schematic diagram illustrating an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. For example... Figure 4 As shown, this embodiment provides a computer-readable storage medium 600, on which a computer program 611 is stored. When the computer program 611 is executed by a processor, it performs the following steps:

[0072] S100: Based on the trajectory data sampled multiple times, the ICP algorithm is used to register map fragments and cut them into tiles.

[0073] S200 rasterizes the tiled map fragments and generates rasterized trajectory density based on the trajectory direction;

[0074] S300 generates a road vector based on the grid and trajectory direction and establishes a correlation with the original trajectory data;

[0075] S400: Fit a probability model based on the density when generating road vectors and obtain the posterior distribution by sampling; calculate the confidence value of the trajectory based on the posterior distribution.

[0076] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0077] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0078] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0079] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0080] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0081] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0082] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for Bayesian trajectory confidence value estimation based on crowd-sourced map aggregation, characterized in that, The application relates to a method for generating a road vector based on trajectory data. The method comprises the following steps: According to the trajectory data of multiple samplings, the map fragments are registered and tiled by using an ICP algorithm; The tiled map fragments are rasterized, and the rasterized trajectory density is generated according to the trajectory direction; Road vectors are formed according to the raster and the trajectory direction, and an associated relationship is established with the original trajectory data; A probability model is fitted according to the density when the road vectors are generated, and a posterior distribution is obtained by sampling; the confidence value of the trajectory is calculated according to the posterior distribution; The rasterization of the tiled map fragments and the generation of the rasterized trajectory density according to the trajectory direction comprise the following steps: The tiled map fragments are rasterized, and each grid represents a pixel point; The rasterized map fragments are divided into four layers according to the trajectory direction, and the four layers correspond to the southwest-to-northeast direction, the northeast-to-southwest direction, the northwest-to-southeast direction and the southeast-to-northwest direction respectively; each layer only contains trajectory data with a matched direction; In each layer, the trajectory pixel points contained in the direction are stacked one by one to form the rasterized trajectory density; The formation of the road vectors according to the raster and the trajectory direction and the establishment of the associated relationship with the original trajectory data comprise the following steps: For any layer, the pixel point P with the maximum trajectory density in the layer is taken; Taking the pixel point P as a starting point, a pixel point P' with the minimum trajectory density difference with the starting point is found in the direction corresponding to the layer; then, in the direction of P->P', a pixel point with the minimum trajectory density difference with the previous pixel point is found pixel by pixel until the edge of the map fragment, the pixel with a density of 0 or the pixel with a density less than a preset threshold B, and a first set is generated; Taking the pixel point P as a starting point, a pixel point P'' with the minimum trajectory density difference with the starting point is found in the direction opposite to the direction corresponding to the layer; then, in the direction of P->P'', a pixel point with the minimum trajectory density difference with the previous pixel point is found pixel by pixel until the edge of the map fragment, the pixel with a density of 0 or the pixel with a density less than a preset threshold B, and a second set is generated; In the direction corresponding to the layer, the pixel points in the first set and the second set are connected to generate a road vector; 2. The method of claim 1, wherein, The trajectory associated with the road vector is found by using a KD-tree space index method, and an associated relationship with the trajectory ID is established. The registration of the map fragments by using the ICP algorithm comprises the following steps: An initial rotation matrix is set as a unit matrix R0=I, and the nearest point set in any two groups of trajectory data is found according to a KD-tree space index; Based on the nearest point set, a rotation matrix R* is solved by using a least square method; any one of the two groups of trajectory data is rotated by using the rotation matrix R*; the nearest point set in the two groups of trajectory data is found again according to the KD-tree space index; the step is repeated until the rotation and translation amount of the rotation matrix R* is less than a preset threshold, and then the iteration is stopped, and the registration of the map fragments is completed; 3. The method of claim 1, wherein, The registered map fragments are tiled according to a specified proportion. The rasterization of the tiled map fragments and the generation of the rasterized trajectory density according to the trajectory direction further comprise the following steps: For each layer, the rasterized map fragments are smoothed by a preset size of convolutional layer.

4. The method of claim 1, wherein, According to the density fitting probability model when generating the road vector, the posterior distribution is obtained by sampling, and the confidence value of the trajectory is calculated according to the posterior distribution, including: According to the association relationship between the road vector and the trajectory, the number of trajectories associated with each road vector is counted, the trajectory number is fitted as a Poisson distribution as a likelihood, and a gamma distribution is taken as a conjugate prior distribution, and the hyperparameter of the gamma distribution is sampled from the non-informative prior distribution or estimated from the data; Then sample the posterior distribution according to Markov Monte Carlo or other Bayesian methods, and obtain the confidence value by calculating the empirical cumulative distribution. 5.A device for Bayesian trajectory confidence value estimation based on crowdsourced map aggregation, comprising: Including: The trajectory position registration module uses the ICP algorithm to register and cut the tile of the map fragments according to the multiple sampled trajectory data; The hierarchical rasterization module rasterizes the tile of the map fragments and generates the rasterized trajectory density according to the trajectory direction; The vector trajectory association module forms the road vector according to the grid and the trajectory direction, and establishes the association relationship with the original trajectory data; The confidence value calculation module fits the probability model according to the density when generating the road vector, obtains the posterior distribution by sampling, and calculates the confidence value of the trajectory according to the posterior distribution; The rasterization of the tile of the map fragments includes: Rasterize the tile of the map fragments, and each grid represents a pixel point; According to the trajectory direction, the rasterized map fragments are divided into four layers, and the four layers correspond to the southwest to northeast direction, the northeast to southwest direction, the northwest to southeast direction, and the southeast to northwest direction, respectively, and each layer only contains trajectory data with the same direction as the corresponding layer; In each layer, the trajectory pixel points contained in the direction are overlaid one by one to form the rasterized trajectory density. The vector trajectory association module forms the road vector according to the grid and the trajectory direction, and establishes the association relationship with the original trajectory data; For any layer, take the pixel point P with the maximum trajectory density in the layer as the starting point; Take the pixel point P as the starting point, find the pixel point P' with the minimum trajectory density difference with the starting point in the direction corresponding to the layer, then find the pixel point with the minimum trajectory density difference with the previous pixel point in the direction of P→P', until the edge of the map fragment, the pixel with a density of 0 or a density less than a preset threshold B, to generate a first set; Take the pixel point P as the starting point, find the pixel point P'' with the minimum trajectory density difference with the starting point in the direction opposite to the direction corresponding to the layer, then find the pixel point with the minimum trajectory density difference with the previous pixel point in the direction of P→P'', until the edge of the map fragment, the pixel with a density of 0 or a density less than a preset threshold B, to generate a second set; Connect the pixel points in the first set and the second set in the direction corresponding to the layer to generate the road vector; Find the trajectory associated with the road vector by using the KD-tree space index method and establish the association relationship with the trajectory ID.

6. An electronic device, comprising: Including: Memory for storing computer software programs; A processor is configured to read and execute the computer software program, thereby implementing the method of claim 1-4.

7. A non-transitory computer-readable storage medium, comprising: The storage medium stores a computer software program, and the computer software program, when executed by a processor, implements the method of claim 1-4.

Images