Road state change detection method and device and electronic equipment

Abstract

The embodiment of the invention discloses a road state change detection method and device and electronic equipment, and the method comprises the steps: obtaining a first track thermodynamic diagram of at least one pre-divided geographic block in a first historical period and a second track thermodynamic diagram of at least one pre-divided geographic block in a second historical period; the first track thermodynamic diagram and the second track thermodynamic diagram of each geographic block are input into a pre-trained change detection model for difference analysis processing, road change information including road change positions and corresponding to each geographic block is obtained, and road state change information is determined according to historical tracks corresponding to the road change positions, the road state change information comprises a change road mark, a road change state and / or a road driving direction. Therefore, track difference analysis can be carried out based on the track thermodynamic diagrams in different periods and the pre-trained change detection model, accurate acquisition of road state change information is realized, and the reliability of road change detection is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and more specifically, to a road state change detection method, device and electronic equipment. Background Art

[0002] With the widespread application of map products, the requirements for map quality are becoming increasingly higher. It is crucial to accurately identify road change information in order to update map data. Summary of the invention

[0003] In view of this, an embodiment of the present invention provides a road state change detection method, device and electronic device, which perform trajectory difference analysis based on trajectory heat maps of different periods and a pre-trained change detection model, thereby achieving accurate acquisition of road state change information and improving the reliability of road change detection.

[0004] In a first aspect, an embodiment of the present invention provides a method for detecting a road state change, the method comprising:

[0005] Obtain a first trajectory heat map of at least one pre-divided geographic block in a first historical period and a second trajectory heat map in a second historical period;

[0006] Inputting the first trajectory heat map and the second trajectory heat map of each of the geographic blocks into a pre-trained change detection model for difference analysis processing, and obtaining road change information corresponding to each of the geographic blocks, wherein the road change information includes a road change state and / or a road change position;

[0007] According to the historical track corresponding to the road change position, the road state change information of the corresponding geographical block is determined, and the road state change information includes a changed road identification, a road change attribute and / or a road driving direction.

[0008] In a second aspect, an embodiment of the present invention provides a road state change detection device, the device comprising:

[0009] A heat map acquisition unit is configured to acquire a first trajectory heat map of at least one pre-divided geographic block in a first historical period and a second trajectory heat map in a second historical period;

[0010] a processing unit configured to input the first trajectory heat map and the second trajectory heat map of each of the geographic blocks into a pre-trained change detection model for difference analysis processing, and obtain road change information corresponding to each of the geographic blocks, wherein the road change information includes a road change state and / or a road change position;

[0011] The information determination unit is configured to determine the road state change information of the corresponding geographical block according to the historical trajectory corresponding to the road change position, and the road state change information includes the changed road identification, road change attributes and / or road driving direction.

[0012] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a memory and a processor, wherein the memory is used to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method as described above.

[0013] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described above is implemented.

[0014] In a fifth aspect, an embodiment of the present invention provides a computer program product, which, when executed on a computer, enables the computer to execute the method described above.

[0015] The embodiment of the present invention obtains the first trajectory heat map of at least one pre-divided geographic block in the first historical period and the second trajectory heat map in the second historical period, and inputs the first trajectory heat map and the second trajectory heat map of each geographic block into a pre-trained change detection model for difference analysis processing, thereby obtaining the road change information including the road change position corresponding to each geographic block, and determining the road state change information according to the historical trajectory corresponding to the road change position, wherein the road state change information includes the changed road sign, the road change state and / or the road driving direction. Therefore, the present embodiment can perform trajectory difference analysis based on trajectory heat maps of different periods and a pre-trained change detection model, thereby achieving accurate acquisition of road state change information and improving the reliability of road change detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0017] Figure 1 is a flow chart of a road state change detection method according to an embodiment of the present invention;

[0018] Figure 2 is a flow chart of a method for obtaining a geographic block according to an embodiment of the present invention;

[0019] Figure 3 This is a schematic diagram of geographical block division according to an embodiment of the present invention;

[0020] Figure 4is another schematic diagram of geographical block division according to an embodiment of the present invention;

[0021] Figure 5 is a flow chart of a method for obtaining a trajectory heat map according to an embodiment of the present invention;

[0022] Figure 6 is a schematic diagram of a trajectory heat map generation process according to an embodiment of the present invention;

[0023] Figure 7 is a flow chart of the processing process of the change detection model of an embodiment of the present invention;

[0024] Figure 8 is a schematic diagram of a change detection model according to an embodiment of the present invention;

[0025] Fig. 9 is a schematic diagram of an encoding unit of an encoder in a change detection model according to an embodiment of the present invention;

[0026] Fig.10 is a schematic diagram of a road state change detection process according to an embodiment of the present invention;

[0027] Fig.11 is a schematic diagram of a road state change detection device according to an embodiment of the present invention;

[0028] Fig.12 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The present application is described below based on embodiments, but the present application is not limited to these embodiments. In the detailed description of the present application below, some specific details are described in detail. It is possible for those skilled in the art to fully understand the present application without the description of these details. In order to avoid confusing the essence of the present application, known methods, processes, flows, components and circuits are not described in detail.

[0030] In addition, persons of ordinary skill in the art will appreciate that the drawings provided herein are for illustration purposes and are not necessarily drawn to scale.

[0031] Unless the context clearly requires otherwise, the words "include", "comprising" and similar words throughout the application should be interpreted as including rather than exclusive or exhaustive; that is, the meaning is "including but not limited to".

[0032] In the description of this application, it should be understood that the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" is two or more.

[0033] The solutions described in this specification and in the examples, if they involve the processing of personal information, will be processed on the premise of having a legal basis (such as obtaining the consent of the subject of personal information, or being necessary for the performance of a contract, etc.), and will only be processed within the scope of regulations or agreements. If a user refuses to process personal information other than the necessary information for basic functions, it will not affect the user's use of basic functions.

[0034] At present, mining new roads based on trajectories is mainly based on extracting the full road network using trajectories, and then performing differential operations with existing maps to determine the abnormal location. However, there are two obvious problems with this process. One is that the process link is long, and executing the full process requires a lot of computing and storage resources, and it takes a long time to execute the mining task. The second is that in complex areas with dense parallel roads, differential operations are difficult. The road matching problem of parallel roads has always been a pain point in the industry. In actual trajectory matching, it often relies on a large amount of upstream and downstream information to make accurate judgments, but this type of information cannot be used when performing differentials on maps. It can only be judged by comparing the generated road network with the existing road network, and the reliability is low. When mining road closures based on trajectories, it is also necessary to bind the trajectory to the existing road network first. In complex scenarios, incorrect binding is more common, which makes the reliability of road closure detection also low. Therefore, the embodiment of the present invention provides a road state change detection method, device and electronic device, which performs trajectory difference analysis based on trajectory heat maps of different periods and pre-trained change detection models, realizes accurate acquisition of road state change information, and improves the reliability of road change detection.

[0035] Figure 1 FIG. 1 is a flow chart of a method for detecting a road state change according to an embodiment of the present invention. Figure 1 As shown, the road state change detection method of the embodiment of the present invention includes the following steps:

[0036] Step S110 , obtaining a first trajectory heat map of at least one pre-divided geographic block in a first historical period and a second trajectory heat map in a second historical period.

[0037] Figure 2 is a flow chart of a method for obtaining a geographic block according to an embodiment of the present invention. In an optional implementation, the geographic block in this embodiment is determined based on a plurality of pre-divided tiles. Specifically, Figure 2 As shown, the method for obtaining a geographic block in this embodiment includes the following steps:

[0038] Step S210 , dividing a predetermined geographical area into a plurality of tiles based on a preset zoom level.

[0039] A tile refers to a square image that is cut into several rows and columns of a map within a certain range according to a certain size and format based on a predetermined zoom level or scale. The higher the zoom level, the more tiles are obtained by dividing the corresponding geographical area, and the smaller the geographical range represented by each tile. In this embodiment, the zoom level of tile division can be determined according to specific needs. For example, 17 tile units can be used, that is, the zoom level is 17. It should be understood that this embodiment is not limited to this.

[0040] Further optionally, the present embodiment may also determine the level of tile division according to the image size adapted to be processed by the adopted change detection model. For example, the image size processed by the change detection model is a*b, where the area represented by each pixel unit is (1 meter * 1 meter), the corresponding zoom level is used for tile division, so that the obtained tile size is approximately (a meter * b meter), or slightly larger than (a meter * b meter).

[0041] Step S220, determining corresponding geographical blocks according to each tile.

[0042] Figure 3 is a schematic diagram of a geographical block division of an embodiment of the present invention. In an optional implementation, this embodiment determines each tile as a corresponding geographical block. Figure 3 As shown, for the geographical area P, tile division is performed according to a preset zoom level to obtain a plurality of tiles w1, and each tile w1 is respectively determined as a geographical block p1.

[0043] Figure 4 is another schematic diagram of geographic block division according to an embodiment of the present invention. In another optional implementation, the present embodiment expands each tile by a predetermined range to obtain each geographic block corresponding to each tile, thereby avoiding the loss of edge information of the geographic block, further improving the accuracy of road state change detection, and further improving the quality of map data updated based on the detected road state change information. It should be understood that the present embodiment does not limit the predetermined range of tile expansion, which can be set according to the specific application scenario.

[0044] like Figure 4 As shown, for the geographic area P, the tiles are divided according to the preset zoom level to obtain a plurality of tiles w1, and the tiles w1 are expanded to a predetermined range to obtain the corresponding geographic block p2 ( Figure 4As an example, other geographic blocks are indicated by dotted lines. Thus, adjacent geographic blocks p2 have a certain overlap range, which avoids the edge effect of the division of geographic blocks, that is, the loss of edge information after the division process leads to discontinuous trajectories or roads, and further improves the accuracy of road state change detection, thereby improving the quality of map data updated based on the detected road state change information.

[0045] It should be understood that this embodiment does not limit the method of dividing geographic blocks. In other optional implementations, this embodiment can also use, for example, a GeoHash-based block division method to divide blocks, which will not be explained in detail one by one.

[0046] In an optional implementation, the first historical period and the second historical period have the same duration, for example, both are 1 day or multiple days, or multiple hours. In other optional implementations, the durations of the first historical period and the second historical period may also be different, and this embodiment does not limit this. Further, the second historical period may be a period closest to the current time, and the first historical period is a period before the second historical period. That is, if the road state changes, the first historical period may represent the period before the road change, and the second historical period may represent the period after the road change.

[0047] In this embodiment, taking the first historical period and the second historical period as 1 day, if the current time is 2024.10.03, 00:00:00, the second historical period may be 2024.10.02, 00:00:00-2024.10.03, 00:00:00, and the first historical period may be 2024.10.01, 00:00:00-2024.10.02, 00:00:00.

[0048] In an optional implementation, this embodiment obtains the first trajectory set and the second trajectory set corresponding to the first historical period and the second historical period in each geographic block, and obtains the corresponding first trajectory heat map and second trajectory heat map based on the first trajectory set and the second trajectory set. Further, this embodiment preprocesses the first trajectory set and the second trajectory set, and obtains the corresponding first trajectory heat map and second trajectory heat map based on the preprocessed first trajectory set and the second trajectory set. The trajectory heat map is used to characterize the trajectory density on the road in the geographic block.

[0049] Figure 5 is a flow chart of a method for obtaining a trajectory heat map according to an embodiment of the present invention. Figure 5 As shown, the trajectory heat map acquisition method of the embodiment of the present invention includes the following steps:

[0050] Step S111 , obtaining a first trajectory set generated by each geographical block in a first historical period and a second trajectory set generated in a second historical period.

[0051] Among them, for the geographic block p, this embodiment obtains the historical trajectories of each road passing through the geographic block p in the first historical period to obtain a first trajectory set, and obtains the historical trajectories of each road passing through the geographic block p in the second historical period to obtain a second trajectory set.

[0052] Step S112, pre-processing the historical trajectories in each first trajectory set and each second trajectory set to remove the historical trajectories that do not meet the predetermined conditions. The predetermined conditions are determined according to the trajectory accuracy, trajectory length, and / or trajectory point speed in the trajectory. That is, this embodiment can remove historical trajectories with low trajectory accuracy, short trajectory length, and abnormal trajectory point speed in the geographic block through pre-processing to avoid interference of such trajectories with road state change detection, thereby further improving the accuracy of road state change detection.

[0053] Step S113: determining the first trajectory heat map of each corresponding geographic block based on each preprocessed first trajectory set.

[0054] Step S114: determining the second trajectory heat map of each corresponding geographic block based on each preprocessed second trajectory set.

[0055] It should be understood that this embodiment does not limit the execution order of the above steps S113 and S114, which may be executed in parallel or sequentially.

[0056] Among them, the size of the first trajectory heat map and the second trajectory heat map is determined based on the size of the corresponding geographic block and the range represented by the pixel unit of the image. Among them, each geographic block has a corresponding picture, and the picture size can be determined according to the size of the area represented by the pixel unit in the picture. For example, the size of the area represented by each pixel unit can be 1m*1m. It should be understood that this embodiment is not limited to this, and it can be limited according to actual application conditions. Therefore, this embodiment can form the corresponding first trajectory heat map and second trajectory heat map based on the pictures corresponding to each geographic block and the trajectory sets of different periods.

[0057] Further optionally, the trajectory heat map of this embodiment can be a schematic diagram showing the trajectory density on the map data of the geographic block, so that the trajectory heat map has relatively rich information. In another optional implementation, the trajectory heat map of this embodiment can be a trajectory grid map, that is, a trajectory heat map after rasterization, and the road state change is determined only based on the trajectory differences in different historical periods. This can avoid the interference of map data in the geographic block, while reducing the computational complexity of data processing and improving detection efficiency. Among them, the raster map is a regular grid image composed of pixels, each pixel represents the brightness value of the point, and the value range is between 0 and 255.

[0058] Figure 6 Schematic diagram of the trajectory heat map generation process of the embodiment of the present invention. Figure 6 As shown, taking the generation process of the first trajectory heat map corresponding to the geographic block p as an example, this embodiment first obtains historical trajectories passing through the roads in the geographic block p during the first historical period to obtain a first trajectory set, preprocesses the first trajectory set to remove historical trajectories with low trajectory accuracy, short trajectory length, or abnormal trajectory point speed in the first trajectory set, obtains the preprocessed first trajectory set, and generates an initial trajectory heat map G1 of the geographic block p based on the first trajectory set. It should be understood that this embodiment can use the initial trajectory heat map G1 as the first trajectory heat map corresponding to the geographic block p.

[0059] Furthermore, in this embodiment, the initial trajectory heat map G1 may be rasterized to obtain a trajectory grid map G2, and the trajectory grid map G2 is used as the first trajectory heat map corresponding to the geographic block p.

[0060] It should be understood that the generation process of the second trajectory heat map corresponding to the geographic block p is similar to the generation process of the first trajectory heat map, and will not be described in detail here.

[0061] Step S120, input the first trajectory heat map and the second trajectory heat map of each geographic block into the pre-trained change detection model for difference analysis processing, and obtain the road change information corresponding to each geographic block. The road change information includes the road change state and / or the road change position. The road change state may include road change and no change. When the road change state is road change, the road change information also includes the road change position. It should be understood that the road change information of this embodiment may also only include the road change position information. When the road change state is no road change, the road change position information is empty.

[0062] In an optional implementation, the change detection model of this embodiment is a model based on the Transformer architecture. It should be understood that this embodiment does not limit the basic architecture of the change detection model, which can extract the feature information of the first trajectory heat map and the second trajectory heat map and calculate and analyze the feature difference between the two. Other basic model architectures, such as model architectures based on deep learning (CNN, FCN, U-net, etc. and their variant architectures) can be applied to this embodiment.

[0063] Furthermore, the change detection model of this embodiment is a ChangeFormer model based on the Transformer architecture.

[0064] Furthermore, the change detection model of this embodiment includes a first encoder, a second encoder, a difference fusion module and a decoder.

[0065] Figure 7 is a flow chart of the processing process of the change detection model of the embodiment of the present invention. In an optional implementation, the change detection model of the present embodiment performs trajectory difference analysis processing through the following steps:

[0066] Step S121: extract features from the first trajectory heat map by using a first encoder to obtain first trajectory features.

[0067] In an optional implementation, the first encoder includes multiple layers of first encoding units, and the first trajectory features include multi-scale first features. That is, the multiple layers of first encoding units extract features of the first trajectory heat map in sequence, and output first features of corresponding scales respectively. It should be understood that this embodiment does not limit the number of first encoding units in the first encoder, which can be comprehensively considered based on model testing effects, model processing efficiency, and computer resource consumption in actual applications.

[0068] Furthermore, each first encoding unit is a Transformer-based encoding unit, including a multi-head self-attention subunit and a position encoding subunit. Optionally, the first encoding unit adopts the following structure:

[0069] F out =MLP(GELU(Conv2D 3×3 (MLP(F in ))))+F in

[0070] Among them, F in Represents the self-attention result. MLP (Multilayer Perceptron) is a module that integrates position encoding, that is, the position encoding subunit mentioned above. Conv2D 3×3A 3×3 2D convolution kernel is used to perform convolution operations to extract trajectory features. GELU (Gaussian Error Linear Unit) is an activation function.

[0071] It should be understood that the structure of the above-mentioned first encoding unit is merely exemplary. This embodiment does not limit the specific structure of the feature extraction, position encoding and self-attention sub-units adopted therein, as long as it can achieve the corresponding functions.

[0072] Step S122: extract features from the second trajectory heat map by using a second encoder to obtain second trajectory features.

[0073] In an optional implementation, the second encoder includes multiple layers of second encoding units, and the second trajectory features include multi-scale second features. That is, the multiple layers of second encoding units extract features of the second trajectory heat map in sequence, and output second features of corresponding scales respectively. It should be understood that this embodiment does not limit the number of second encoding units in the second encoder, which can be comprehensively considered based on the model testing effect, model processing efficiency, and computer resource consumption in actual applications.

[0074] Furthermore, the first encoding unit in the second encoder corresponds to the first encoding unit in the first encoder one by one, that is, the trajectory difference feature at the corresponding scale is calculated by using the first feature and the second feature at the same scale.

[0075] Furthermore, the structure of the second encoding unit in this embodiment is the same as or similar to that of the first encoding unit. For details, please refer to the above-mentioned structure of the first encoding unit, which will not be described in detail here.

[0076] Step S123: Process the first trajectory feature and the second trajectory feature through a difference fusion module to obtain a trajectory difference feature.

[0077] In an optional implementation, the difference fusion module includes a plurality of layers of difference fusion units, and the trajectory difference features include trajectory difference features at different scales.

[0078] In this embodiment, the outputs of the first encoding unit and the second encoding unit that output features of the same scale are connected to the corresponding difference fusion unit, and the difference fusion unit receives the first feature and the second feature output by the corresponding first encoding unit and the second encoding unit to perform difference feature extraction to obtain trajectory difference features at the corresponding scale.

[0079] In an optional implementation, the difference fusion unit includes a convolution layer, a Relu layer, and a BN layer (BatchNormalization Layer), and the specific structure is:

[0080]

[0081] Among them, i is the difference fusion unit of the i-th layer, is the trajectory difference feature at the corresponding scale output by the difference fusion unit of the i-th layer, is the output of the first encoding unit of the i-th layer, is the output of the second encoding unit of the i-th layer, Characterization feature map and feature map The feature concatenation result of Conv2D 3×3 A 3×3 two-dimensional convolution kernel is used for convolution operation to extract trajectory features. GELU (Gaussian Error Linear Unit) is an activation function, and BN is a batch normalization layer.

[0082] It should be understood that the structure of the above-mentioned difference fusion unit is merely exemplary, and this embodiment does not limit the specific structure of the difference feature extraction adopted by it, as long as it can achieve the corresponding function.

[0083] Step S124, the decoder analyzes the trajectory difference features to obtain the road change information of the corresponding geographical block. Further, the decoder is used to fuse the trajectory difference features at different scales to generate a difference feature change map, which includes the road change information of the corresponding geographical block.

[0084] Furthermore, the difference feature change map may be a binary change map (CM), that is, the change detection result is represented in a binary form, wherein the change area and the non-change area are distinguished by different pixel values, thereby making it easier to distinguish the trajectory changes in different periods. It should be understood that the output of the decoder may not be a binary change map, and it only needs to be able to characterize the road change state and the change position. This embodiment does not limit the form of its output, and this embodiment can set the model based on demand.

[0085] Optionally, the decoder of this embodiment can be a decoder based on a lightweight MLP to fuse trajectory difference features at different scales and predict road change information. It should be understood that this embodiment does not limit the selection of the decoder, and it only needs to be able to achieve the fusion and analysis of difference features.

[0086] Figure 8 is a schematic diagram of a change detection model according to an embodiment of the present invention. It should be understood that Figure 8 The change detection model shown is only exemplary, and the present embodiment is not limited thereto. Figure 8As shown, this embodiment obtains the first trajectory heat map Pre-G and the second trajectory heat map Post-G based on any of the above implementation methods.

[0087] The change detection model of this embodiment includes a first encoder 81, a second encoder 82, a difference fusion module 83 and a decoder 84. Optionally, the first encoder 81 and the second encoder 82 of this embodiment have the same structure, and a weight sharing method can be adopted to reduce the number of parameters of the model. It should be understood that the weights of the first encoder 81 and the second encoder 82 may not be shared, and this embodiment does not limit this.

[0088] Among them, the first encoder 81 includes first encoding units 811-814. Further, the first encoder 81 also includes a plurality of downsampling units to perform downsampling processing before the first encoding unit performs feature processing. The second encoder 82 includes second encoding units 821-824. Further, the second encoder 82 also includes a plurality of downsampling units to perform downsampling processing before the second encoding unit performs feature processing. The difference fusion module 83 includes difference fusion units 831-834. In this embodiment, the output of the first encoding unit 811 and the output of the second encoding unit 821 are both connected to the difference fusion unit 831, the output of the first encoding unit 812 and the output of the second encoding unit 822 are both connected to the difference fusion unit 832, the output of the first encoding unit 813 and the output of the second encoding unit 823 are both connected to the difference fusion unit 833, and the output of the first encoding unit 814 and the output of the second encoding unit 824 are both connected to the difference fusion unit 834.

[0089] Further, in this embodiment, after receiving the input first trajectory heat map Pre-G, the first encoder 81 downsamples it and performs feature processing through the first encoding unit 811 to obtain the first feature The first encoder 81 performs the first feature After downsampling, it is input to the first encoding unit 812 for feature processing to obtain the first feature The first encoder 81 performs the second feature After downsampling, it is input to the first encoding unit 813 for feature processing to obtain the first feature The first encoder 81 performs the first feature After downsampling, it is input to the first encoding unit 814 for feature processing to obtain the first feature

[0090] After receiving the input second trajectory heat map Post-G, the second encoder 82 downsamples it and performs feature processing through the second encoding unit 821 to obtain the second feature The second encoder 82 performs the second feature After downsampling, it is input to the second encoding unit 822 for feature processing to obtain the second feature The second encoder 82 performs the second feature After downsampling, it is input to the second encoding unit 823 for feature processing to obtain the second feature The second encoder 82 performs the second feature After downsampling, it is input to the second encoding unit 824 for feature processing to obtain the second feature

[0091] Fig. 9 is a schematic diagram of the encoding unit of the encoder in the change detection model of the embodiment of the present invention. In an optional implementation, the structures of the first encoding units 811-814 in the first encoder 81 and the second encoding units 821-824 in the second encoder 82 of the present embodiment are similar, such as Fig. 9 As shown, the structure of the encoding unit 9 (i.e., the first encoding unit or the second encoding unit mentioned above) includes N modules consisting of a multi-head self-attention sub-unit (Multi-Head Self-Attention) 91 and an address encoding sub-unit (Positional Encoding) 92. It should be understood that this embodiment does not limit the number of N, which can be set according to the specific situation of data processing in actual applications.

[0092] Among them, the multi-head self-attention sub-unit 91 captures different information of the input sequence in different subspaces by executing multiple heads (i.e., sub-attention operations) in parallel, generates a sequence s1, and improves the expression ability of the model. The address encoding sub-unit 92 generates a sequence s2 by encoding the position information into a vector, and controls the addition of the sequence s2 to the sequence s1 output by the multi-head self-attention sub-unit 91, so that the model can perceive the position information in the sequence, and then generate the corresponding first feature or second feature.

[0093] Further, the address encoding subunit 92 is implemented by sequentially connecting MLP 921, depth-wise convolution subunit 922 and MLP 923. It should be understood that this embodiment does not limit the specific implementation of the address encoding subunit 92, as long as it can implement address encoding.

[0094] Furthermore, the difference fusion unit 831 receives the first feature And the second feature Perform differential information extraction to obtain trajectory difference features at the corresponding scale The difference fusion unit 832 receives the first feature And the second feature Perform differential information extraction to obtain trajectory difference features at the corresponding scale The difference fusion unit 833 receives the first feature And the second feature Perform differential information extraction to obtain trajectory difference features at the corresponding scale The difference fusion unit 834 receives the first feature And the second feature Perform differential information extraction to obtain trajectory difference features at the corresponding scale

[0095] Furthermore, the decoder 84 receives and fuses the trajectory difference features at different scales Trajectory difference characteristics Trajectory difference characteristics and trajectory difference characteristics Obtain the difference feature change graph G0. In this embodiment, the difference feature change graph G0 is taken as a binary graph as an example.

[0096] Specifically, the decoder 84 receives the trajectory difference features at different scales through MLP841. Trajectory difference characteristics Trajectory difference characteristics and trajectory difference characteristics Processing is performed, and the processed features are upsampled and spliced, and the spliced ​​features are input into MLP842 for processing, and then upsampled by upsampling module 845 and input into classifier 846, and the trajectory difference information is classified by classifier 846 to obtain the difference feature change graph G0. Optionally, the classifier 846 determines the difference feature greater than the predetermined difference threshold as a change (for example, determining its value to be 1), and determines the difference feature not greater than the predetermined difference threshold as unchanged (for example, determining its value to be 0), so as to obtain a binary type difference feature change graph G0. It should be understood that the present embodiment does not limit the specific structure of the decoder, and it can realize the fusion and analysis of trajectory difference features at different scales. In other optional implementations, the decoder 84 can also directly output the change value corresponding to each pixel unit to obtain the difference feature change graph G0, and the present embodiment does not limit this.

[0097] In this embodiment, the difference feature extraction of the trajectory heat map of different periods can be performed through the change detection model to obtain the difference feature change map, and then obtain the road change information in each geographical block.

[0098] Furthermore, the change detection model of the present embodiment can be obtained by training based on a pre-sampled sample set. The sample set includes a plurality of samples, and each sample includes a first sample trajectory and a second sample trajectory of a corresponding geographical block at different periods. Optionally, the samples in the sample set may have corresponding road change labels (labeled with road change status, road change position and / or road change attributes, etc.). In other optional implementations, the sample set may include some samples with road change labels or include samples without road change labels, and the present embodiment does not limit this. Furthermore, the samples in the sample set are input into the change detection model for processing, and the weights of each part in the change detection model are adjusted based on the output of the model to obtain a trained change detection model.

[0099] Further optionally, the present embodiment may obtain a sample set through a field driving test, may obtain a sample set through simulated driving using relevant simulation software, or may obtain a sample set based on collectible historical driving trajectories of the vehicle. The present embodiment does not limit the method for obtaining the sample set.

[0100] Step S130, according to the historical track corresponding to the road change position, determine the road state change information of the corresponding geographical block. The road state change information includes the changed road identification, the road change attribute and / or the road driving direction. The road change attribute can also be whether the road is a newly added road or a closed road (i.e., changed from a passable state to a closed and impassable state).

[0101] In this embodiment, for the geographical block with road changes determined by the change detection model, the difference feature change map of the geographical block is topologically restored, that is, by extracting the skeleton line result of the road change position, the position of the road change position in the picture is converted into longitude and latitude position information. Among them, the skeleton line is used to represent the position, direction and attributes of the linear objects in the map, such as the road position and direction, road type, width and grade, etc.

[0102] In this embodiment, the historical tracks passing through the road change position in the corresponding first track set and the second track set are bound to the corresponding road to determine the road change attribute and the road travel direction of the road change position. It is easy to understand that when the historical track is bound to the corresponding road, if the historical track on the road is from existence to non-existence, the road change attribute is a closed road, and if the historical track on the road is from non-existence to existence, the road change attribute is a newly added road. In addition, the road travel direction can be determined by the direction of the track.

[0103] In an optional implementation, when a road change in a geographic block is detected by a change detection model, the present embodiment may also send the road state change information of the geographic block where the road change occurs to the audit terminal, so as to audit the road state change information of the geographic block where the road change occurs. Further optionally, the present embodiment may adopt manual audit or machine audit, for example, after the audit terminal receives the road state change information of the geographic block, the auditor or the machine obtains the historical track or driving recorder image of the geographic block for audit judgment to determine whether the road state change information is accurate, and then determine whether to change the map data based on it.

[0104] In an optional implementation, the present embodiment can also update the map data based on the road status change information of each geographic block to facilitate subsequent accurate navigation. For example, if a geographic block changes, new roads and road signs are added to the map corresponding to the geographic block at the road change location based on the road change attributes and road driving direction, or the corresponding closed roads are deleted, or road closure information is displayed.

[0105] The embodiment of the present invention obtains the first trajectory heat map of at least one pre-divided geographic block in the first historical period and the second trajectory heat map in the second historical period, and inputs the first trajectory heat map and the second trajectory heat map of each geographic block into a pre-trained change detection model for difference analysis processing, thereby obtaining the road change information including the road change position corresponding to each geographic block, and determining the road state change information according to the historical trajectory corresponding to the road change position, wherein the road state change information includes the changed road sign, the road change state and / or the road driving direction. Therefore, the present embodiment can perform trajectory difference analysis based on trajectory heat maps of different periods and a pre-trained change detection model, thereby achieving accurate acquisition of road state change information and improving the reliability of road change detection.

[0106] Fig.10 FIG. 1 is a schematic diagram of a road state change detection process according to an embodiment of the present invention. Fig.10 As shown, for a geographic area P including roads r1, r2 and r3, tile division is performed according to a preset zoom level to obtain a plurality of tiles, and the tiles are expanded to a predetermined range to obtain a corresponding geographic block p. As a result, adjacent geographic blocks p have a certain overlapping range, which avoids the edge effect of the division of geographic blocks, that is, the loss of edge information after the division process, resulting in discontinuous trajectories or roads, and further improves the accuracy of road state change detection, thereby improving the quality of map data updated based on the detected road state change information.

[0107] Furthermore, historical trajectories of each road passing through the geographic block p in a first historical period are obtained to obtain a first trajectory set, and historical trajectories of each road passing through the geographic block p in a second historical period are obtained to obtain a second trajectory set. The historical trajectories in each first trajectory set and the second trajectory set are preprocessed to remove historical trajectories that do not meet predetermined conditions. A first trajectory heat map Pre-G1 of each corresponding geographic block is determined based on each preprocessed first trajectory set, and a second trajectory heat map Post-G1 of each corresponding geographic block is determined based on each preprocessed second trajectory set.

[0108] Furthermore, each first trajectory heat map Pre-G1 and each second trajectory heat map Post-G1 are input into the change detection model 10, and the difference feature change map G' corresponding to each geographic block p is obtained. The geographic block p with road changes is determined according to the difference feature change map G' corresponding to each geographic block p, and the difference feature change map G' of the geographic block p with road changes is topologically restored, that is, by extracting the skeleton line result of the road change position, the position of the road change position in the picture is converted into longitude and latitude position information, and the historical tracks passing through the road change position in the corresponding first trajectory set and the second trajectory set are bound to the corresponding road to determine the road change attributes and road driving direction of the road change position. Therefore, this embodiment can determine the road state change information of each geographic block with road changes, and update the road state change information to the map corresponding to each geographic block to obtain the updated geographic block p'.

[0109] Furthermore, since the geographic block p' is obtained by expanding the predetermined range of the divided tiles, adjacent geographic blocks p' have a certain overlap. Therefore, this embodiment cuts off the corresponding predetermined range on the geographic block p' and splices the adjacent cut geographic blocks p' to obtain the map corresponding to the updated geographic area P. Fig.10 As shown, the updated geographical area P is newly added with road r4, and road r1 is closed.

[0110] The embodiment of the present invention obtains the first trajectory heat map of at least one pre-divided geographic block in the first historical period and the second trajectory heat map in the second historical period, and inputs the first trajectory heat map and the second trajectory heat map of each geographic block into a pre-trained change detection model for difference analysis processing, thereby obtaining the road change information including the road change position corresponding to each geographic block, and determining the road state change information according to the historical trajectory corresponding to the road change position, wherein the road state change information includes the changed road sign, the road change state and / or the road driving direction. Therefore, the present embodiment can perform trajectory difference analysis based on trajectory heat maps of different periods and a pre-trained change detection model, thereby achieving accurate acquisition of road state change information and improving the reliability of road change detection.

[0111] Fig.11 Schematic diagram of a road state change detection device according to an embodiment of the present invention. Fig.11 As shown, the road state change detection device 11 of the embodiment of the present invention includes a heat map acquisition unit 111, a processing unit 112 and an information determination unit 113.

[0112] The heat map acquisition unit 111 is configured to acquire a first trajectory heat map of at least one pre-divided geographic block in a first historical period and a second trajectory heat map in a second historical period. The processing unit 112 is configured to input the first trajectory heat map and the second trajectory heat map of each of the geographic blocks into a pre-trained change detection model for difference analysis processing, and acquire the road change information corresponding to each of the geographic blocks, the road change information including the road change state and / or the road change position. The information determination unit 113 is configured to determine the road state change information of the corresponding geographic block based on the historical trajectory corresponding to the road change position, the road state change information including the changed road identification, road change attributes and / or road driving direction. Optionally, the first trajectory heat map and the second trajectory heat map are trajectory grid maps.

[0113] In an optional implementation, the heat map acquisition unit 111 is further configured to perform: acquiring a first trajectory set generated in a first historical period and a second trajectory set generated in a second historical period for each of the geographic blocks, preprocessing the historical trajectories in each of the first trajectory set and the second trajectory set to remove historical trajectories that do not meet a predetermined condition, wherein the predetermined condition is determined according to trajectory accuracy, trajectory length, and / or trajectory point speed in the trajectory, determining the first trajectory heat map of each of the corresponding geographic blocks based on each of the preprocessed first trajectory sets, and determining the second trajectory heat map of each of the corresponding geographic blocks based on each of the preprocessed second trajectory sets.

[0114] In an optional implementation, the change detection model includes a first encoder, a second encoder, a difference fusion module and a decoder. The processing unit 112 is further configured to perform: extracting features from the first trajectory heat map through the first encoder to obtain first trajectory features, extracting features from the second trajectory heat map through the second encoder to obtain second trajectory features, processing the first trajectory features and the second trajectory features through the difference fusion module to obtain trajectory difference features, parsing the trajectory difference features through the decoder to obtain road change information of the corresponding geographic block.

[0115] In an optional implementation, the first encoder includes multiple layers of first encoding units, the first trajectory features include multi-scale first features, the second encoder includes multiple layers of second encoding units, the second trajectory features include multi-scale second features, the difference fusion module includes multiple layers of difference fusion units, and the trajectory difference features include trajectory difference features at different scales.

[0116] In an optional implementation, the state change detection device 11 further includes a tile division unit and a block determination unit. The tile division unit is configured to divide the predetermined geographical area into a plurality of tiles based on a preset zoom level. The block determination unit is configured to determine the corresponding geographical blocks according to the tiles.

[0117] In an optional implementation, the block determination unit is configured to further determine each of the tiles as the corresponding geographical blocks; or to expand each of the tiles by a predetermined range to obtain each of the geographical blocks corresponding to each tile.

[0118] In an optional implementation, the state change detection device 11 further includes a sending unit configured to send the road state change information to a review terminal to review the road state change information.

[0119] In an optional implementation, the state change detection device 11 further includes an updating unit configured to update map data based on the road state change information of each of the geographical blocks.

[0120] The embodiment of the present invention obtains the first trajectory heat map of at least one pre-divided geographic block in the first historical period and the second trajectory heat map in the second historical period, and inputs the first trajectory heat map and the second trajectory heat map of each geographic block into a pre-trained change detection model for difference analysis processing, thereby obtaining the road change information including the road change position corresponding to each geographic block, and determining the road state change information according to the historical trajectory corresponding to the road change position, wherein the road state change information includes the changed road sign, the road change state and / or the road driving direction. Therefore, the present embodiment can perform trajectory difference analysis based on trajectory heat maps of different periods and a pre-trained change detection model, thereby achieving accurate acquisition of road state change information and improving the reliability of road change detection.

[0121] Fig.12 Schematic diagram of an electronic device according to an embodiment of the present invention. Fig.12 As shown, the electronic device 120 is a general data processing device, which includes a general computer hardware structure, which at least includes a processor 121 and a memory 122. The processor 121 and the memory 122 are connected via a bus 123. The memory 122 is suitable for storing instructions or programs executable by the processor 121. The processor 121 can be an independent microprocessor or a collection of one or more microprocessors. Thus, the processor 121 executes the instructions stored in the memory 122, thereby executing the method flow of the embodiment of the present invention as described above to realize the processing of data and the control of other devices. The bus 123 connects the above-mentioned multiple components together, and at the same time connects the above-mentioned components to the display controller 124 and the display device and the input / output (I / O) device 125. The input / output (I / O) device 125 can be a mouse, a keyboard, a modem, a network interface, a touch input device, a somatosensory input device, a printer, and other devices known in the art. Typically, the input / output device 125 is connected to the system through an input / output (I / O) controller 126.

[0122] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, devices (equipment) or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may adopt a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0123] The present application is described with reference to flowcharts of methods, apparatuses (devices) and computer program products according to embodiments of the present application. It should be understood that each process in the flowchart can be implemented by computer program instructions.

[0124] These computer program instructions may be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device that implements the process Figure 1 A function specified in a process or multiple processes.

[0125] These computer program instructions may also be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce the instructions for implementing the process Figure 1 A device that specifies functions in a process or multiple processes.

[0126] Another embodiment of the present invention relates to a non-volatile storage medium for storing a computer-readable program, wherein the computer-readable program is used for a computer to execute part or all of the above method embodiments.

[0127] That is, those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by specifying relevant hardware through a program, and the program is stored in a storage medium, including several instructions for a device (which can be a single-chip microcomputer, chip, etc.) or a processor (processor) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0128] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A road state change detection method, characterized in that: The method comprises: Obtain a first trajectory heat map of at least one pre-divided geographic block in a first historical period and a second trajectory heat map in a second historical period; Inputting the first trajectory heat map and the second trajectory heat map of each of the geographic blocks into a pre-trained change detection model for difference analysis and processing, and obtaining road change information corresponding to each of the geographic blocks, wherein the road change information includes a road change state and / or a road change position; According to the historical track corresponding to the road change position, the road state change information of the corresponding geographical block is determined, and the road state change information includes a changed road identification, a road change attribute and / or a road driving direction.

2. The method according to claim 1, characterized in that The obtaining of a first trajectory heat map of at least one pre-divided geographic block in a first historical period and a second trajectory heat map in a second historical period comprises: Obtain a first trajectory set generated in a first historical period and a second trajectory set generated in a second historical period for each of the geographic blocks; Preprocessing the historical trajectories in each of the first trajectory set and the second trajectory set to remove historical trajectories that do not meet a predetermined condition, wherein the predetermined condition is determined according to trajectory accuracy, trajectory length, and / or trajectory point speed in the trajectory; Determine the first trajectory heat map of each corresponding geographic block based on each preprocessed first trajectory set; The second trajectory heat map of each corresponding geographic block is determined based on each preprocessed second trajectory set.

3. The method according to claim 1, characterized in that: The change detection model includes a first encoder, a second encoder, a difference fusion module and a decoder; The step of inputting the first trajectory heat map and the second trajectory heat map of each of the geographic blocks into a pre-trained change detection model for difference analysis and processing to obtain the road change information corresponding to each of the geographic blocks includes: Extracting features from the first trajectory heat map by using the first encoder to obtain first trajectory features; Performing feature extraction on the second trajectory heat map by using the second encoder to obtain second trajectory features; Processing the first trajectory feature and the second trajectory feature by the difference fusion module to obtain a trajectory difference feature; The track difference feature is analyzed by the decoder to obtain the road change information of the corresponding geographical block.

4. The method according to claim 3, characterized in that The first encoder includes multiple layers of first encoding units, the first trajectory features include multiple-scale first features, the second encoder includes multiple layers of second encoding units, the second trajectory features include multiple-scale second features, the difference fusion module includes multiple layers of difference fusion units, and the trajectory difference features include trajectory difference features at different scales.

5. The method according to claim 1, characterized in that The method further comprises: dividing a predetermined geographic area into a plurality of tiles based on a preset zoom level; The corresponding geographical blocks are determined according to the tiles.

6. The method according to claim 5, characterized in that Determining the corresponding geographical blocks according to the tiles comprises: Determine each of the tiles as the corresponding geographical blocks; or Each tile is expanded by a predetermined range to obtain each geographical block corresponding to each tile.

7. The method according to claim 1, characterized in that The method further comprises: The road state change information is sent to a review terminal to review the road state change information.

8. The method according to claim 1, characterized in that: The method further comprises: Map data is updated based on the road status change information of each of the geographical blocks.

9. The method according to claim 1, characterized in that: The first trajectory heat map and the second trajectory heat map are trajectory grid maps.

10. A road state change detection device, characterized in that: The device comprises: A heat map acquisition unit is configured to acquire a first trajectory heat map of at least one pre-divided geographic block in a first historical period and a second trajectory heat map in a second historical period; a processing unit configured to input the first trajectory heat map and the second trajectory heat map of each of the geographic blocks into a pre-trained change detection model for difference analysis processing, and obtain road change information corresponding to each of the geographic blocks, wherein the road change information includes a road change state and / or a road change position; The information determination unit is configured to determine the road state change information of the corresponding geographical block according to the historical trajectory corresponding to the road change position, and the road state change information includes the changed road identification, road change attributes and / or road driving direction.

11. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

13. A computer program product, characterized in that When the computer program product is run on a computer, the computer is caused to execute the method according to any one of claims 1 to 9.

Images