Map matching method and device based on semantic features, storage medium and product

Abstract

One or more embodiments of the present specification provide a map matching method and device based on semantic features, a storage medium and a product, the method comprising: obtaining trajectory data of a target object and corresponding semantic data, the trajectory data being used for representing a moving trajectory of the target object, and the semantic data being used for representing the moving trajectory of the target object; the semantic data is used for representing road traffic information corresponding to the moving track; determining track features corresponding to the track data, and determining semantic features corresponding to the semantic data; and determining a target path matched with the moving track and the road traffic information in a target map according to the track feature and the semantic feature.

Description

Technical Field

[0001] One or more embodiments of the present specification relate to the field of computer technology, and in particular, to a map matching method, device, storage medium, and product based on semantic features. Background Art

[0002] Map matching technology can accurately compare and correct the movement trajectory of vehicles or mobile devices with the actual road network on the digital map. This technology uses complex algorithm processing to adjust the deviation trajectory points that may be caused by poor positioning signals such as the Global Positioning System (GPS) to the closest actual road and path, thereby greatly improving the accuracy of location information. Map matching is widely used in navigation services, traffic flow analysis, and intelligent transportation systems, helping to provide a more reliable and accurate service experience.

[0003] In related technologies, map matching technology is usually determined based on the geometric spatial relationship between the moving trajectory and the map path, such as the straight-line distance from the trajectory point to the map location point, and the angle between the movement direction and the map path direction. However, when faced with complex urban road networks, the emergence of special roads such as roundabouts, U-shaped bends, and overpasses will complicate the above geometric spatial relationship, thereby affecting the accuracy of map matching. At the same time, due to the obstruction of factors such as buildings and terrain, the signals of positioning devices such as GPS may be interfered, resulting in an increase in the positioning error of the above trajectory points and a decrease in the positioning frequency, further reducing the accuracy of map matching. Summary of the invention

[0004] In view of this, one or more embodiments of this specification provide the following technical solutions:

[0005] According to a first aspect of one or more embodiments of this specification, a map matching method based on semantic features is proposed, the method comprising:

[0006] Acquire trajectory data of a target object and its corresponding semantic data, wherein the trajectory data is used to characterize a moving trajectory of the target object, and the semantic data is used to characterize road traffic information corresponding to the moving trajectory;

[0007] Determining a trajectory feature corresponding to the trajectory data, and determining a semantic feature corresponding to the semantic data;

[0008] A target path in a target map that matches the moving trajectory and the road traffic information is determined according to the trajectory feature and the semantic feature.

[0009] Optionally, the trajectory data includes positioning information of multiple trajectory points; and determining a trajectory feature corresponding to the trajectory data includes:

[0010] Determine a first type of trajectory feature of each trajectory point according to the positioning information of each trajectory point in the trajectory data;

[0011] The second type of trajectory features of each trajectory point are determined according to the positioning information between adjacent trajectory points in the trajectory data, and the second type of trajectory features and the first type of trajectory features are used as the trajectory features.

[0012] Optionally, the first type of trajectory features includes at least any one of the following: positioning features, time features; and / or,

[0013] The second type of trajectory features includes at least any one of the following: speed features, slope features, curvature features, and trajectory quality features.

[0014] Optionally, determining a target path in a target map that matches the moving trajectory and the road traffic information according to the trajectory feature and the semantic feature includes:

[0015] Determining, according to the trajectory features, a plurality of candidate map sections that match the movement trajectory;

[0016] determining a matching score between each candidate map segment and the movement trajectory according to the trajectory feature and the semantic feature, wherein the matching score is positively correlated with a first similarity between the movement trajectory and the corresponding candidate map segment and a second similarity between the road traffic information corresponding to the movement trajectory and the road traffic information in the corresponding candidate map segment;

[0017] The target path is determined according to the matching score and the map data of the target map.

[0018] Optionally, determining a matching score between each candidate map segment and the moving trajectory according to the trajectory feature and the semantic feature includes:

[0019] In a case where the trajectory feature includes at least any one of a positioning feature, a speed feature, a slope feature, and a curvature feature, calculating a feature difference between each candidate map segment and the moving trajectory for the at least any one of the features, and determining the matching score according to the feature difference, wherein the feature difference is negatively correlated with the matching score;

[0020] In the case where the trajectory feature includes a trajectory quality feature, a score weight of the trajectory feature corresponding to the trajectory quality feature is calculated, and the matching score is determined according to the score weight, and the trajectory quality feature is positively correlated with the score weight and the matching score, respectively.

[0021] Optionally, the semantic feature is determined according to semantic data, or the semantic feature is determined according to the trajectory feature and the semantic data; and / or,

[0022] The semantic data is acquired through a road sensor of the target object; and / or, the semantic data is acquired through a reference map.

[0023] Optionally, determining the semantic feature according to the trajectory feature and the semantic data includes:

[0024] A road scene corresponding to the moving trajectory is identified according to the trajectory feature and the semantic data, and the road scene is used as the semantic feature.

[0025] According to a second aspect of one or more embodiments of this specification, an electronic device is proposed, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor implements the steps of the method described in the first aspect by running the executable instructions.

[0026] According to a third aspect of one or more embodiments of the present specification, a computer-readable storage medium is provided, on which computer instructions are stored, and when the instructions are executed by a processor, the steps of the method described in the first aspect are implemented.

[0027] According to a fourth aspect of one or more embodiments of the present specification, a computer program product is provided, comprising a computer program / instruction, which implements the steps of the method described in the first aspect when executed by a processor.

[0028] In the technical solution provided in the present specification, in addition to obtaining trajectory data for characterizing the moving trajectory of the target object, semantic data for characterizing the road traffic information corresponding to the moving trajectory is also obtained. After determining the corresponding trajectory features and semantic features based on the trajectory data and the semantic data, the target path in the target map that matches the moving trajectory can also be determined based on the above trajectory features and the above semantic features. The introduction of semantic data enables the map matching process of the present specification to jointly determine the above target path based on the trajectory features and the semantic features corresponding to the above road traffic information. Since the semantic data does not belong to positioning information, it does not involve the calculation of geometric spatial relationships related to the positioning information, and is less affected by special circumstances such as complex urban road networks. The accuracy of map matching can be effectively improved by combining with trajectory data. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments or prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention, and for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work.

[0030] Figure 1 is a flowchart of a map matching method based on semantic features provided by an exemplary embodiment;

[0031] Figure 2 is a flowchart of a method for acquiring trajectory features provided by an exemplary embodiment;

[0032] Figure 3 is a flowchart of a method for acquiring semantic features provided by an exemplary embodiment;

[0033] Figure 4 is a schematic diagram of a flow chart of determining a target path provided by an exemplary embodiment;

[0034] Figure 5 It is a schematic diagram of a flow chart of calculating a matching score between an alternative map path and a moving trajectory provided by an exemplary embodiment;

[0035] Figure 6 is a schematic structural diagram of an electronic device provided by an exemplary embodiment;

[0036] Figure 7 It is a structural schematic diagram of a map matching device provided by an exemplary embodiment. DETAILED DESCRIPTION

[0037] Here, exemplary embodiments will be described in detail, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are only examples of devices and methods consistent with some aspects of the present invention.

[0038] The terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "the" and "the" used in the present invention and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0039] It should be understood that although the terms first, second, third, etc. may be used in the present invention to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0040] Figure 1 FIG. 1 is a flowchart of a map matching method based on semantic features provided by an exemplary embodiment. Figure 1 As shown, the above method includes:

[0041] Step S102: Acquire trajectory data of a target object and its corresponding semantic data, wherein the trajectory data is used to represent a moving trajectory of the target object, and the semantic data is used to represent road traffic information corresponding to the moving trajectory.

[0042] When it is necessary to determine the actual path of a moving target object in the target map, the trajectory data of the target object and the semantic data corresponding to the trajectory data can be obtained first. The target object can be any device or object with mobile capability, such as a car or a running athlete.

[0043] The above-mentioned trajectory data can be acquired by the target object or a data acquisition device that moves with the target object, such as the GPS positioning sensor of the above-mentioned car, or the smart electronic device that supports positioning function worn by the above-mentioned athlete, etc. Correspondingly, the trajectory data may include positioning information, that is, location information and time information. The above-mentioned location information may specifically include the longitude, latitude and altitude of the target object, and the above-mentioned time information may include a timestamp for acquiring the above-mentioned location information. Obviously, each positioning information can represent the position point of the target object at the corresponding moment, hereinafter referred to as a trajectory point. Then, when the above-mentioned trajectory data contains positioning information of multiple trajectory points, the line segment continuously formed by the above-mentioned multiple trajectory points can be used to represent the movement trajectory of the target object within a time period.

[0044] The above semantic data can be obtained by the target object or a data acquisition device that moves with the target object, such as image or video data captured by a camera in a road sensor configured in a car, or it can be obtained from a reference map configured in the above target object, which can be the above target map or a preset electronic map different from the above target map. In short, this solution can choose any of the above methods to obtain semantic data according to actual conditions, thereby improving the flexibility of obtaining data and solving the problem of low success rate of semantic data acquisition compared to a single acquisition channel. Correspondingly, the above semantic data can refer to road traffic information on the road corresponding to the above moving trajectory, such as one or more traffic elements such as traffic lights, zebra crossings, and traffic signs.

[0045] It should be noted that the correspondence between the above-mentioned trajectory data and the above-mentioned semantic data can be established based on the acquisition time. For example, the above-mentioned correspondence is established when the acquisition time point of the two is not greater than the preset acquisition time. The so-called preset acquisition time can be adjusted according to the degree of delay in the acquisition of trajectory data and semantic data. For example, when the signal positioning strength of the GPS positioning sensor of the above-mentioned car is weak, the above-mentioned preset acquisition time is increased, otherwise the above-mentioned preset acquisition time is reduced.

[0046] Step S104, determining the trajectory features corresponding to the trajectory data, and determining the semantic features corresponding to the semantic data.

[0047] After the data acquisition device acquires the trajectory data and the semantic data, it can be further processed by a corresponding data processing device such as the core processor of the vehicle or the intelligent electronic device or an independent data processing module to extract the trajectory features corresponding to the trajectory data and the semantic features corresponding to the semantic data, which can be used for training models or analysis. The trajectory features can be determined based on the trajectory data, and the semantic features can be determined based on the semantic data, or can be determined based on the determined trajectory features and the semantic data. The acquisition methods of the two can be combined in any way according to actual needs, and this is not limited in this specification.

[0048] Combine the following Figure 2 , the process of determining trajectory features is further explained, such as Figure 2 As shown, the process includes the following steps:

[0049] Step S202: De-noising the trajectory data.

[0050] In one embodiment, the positioning information of some trajectory points contained in the trajectory data may cause data anomalies due to signal interference and other reasons. For example, the difference between the longitude value corresponding to a certain trajectory point and the adjacent trajectory point is greater than a preset anomaly threshold, thereby causing the trajectory point to be determined as an anomaly point. At this time, the trajectory data can be denoised to remove the anomalies or noise in the trajectory data, making the trajectory data smoother and more accurate.

[0051] Specifically, the above-mentioned denoising process can be implemented by smoothing filtering, for example, using a low-pass filter such as a moving average filter, a Kalman filter, etc. to smooth the positioning information of each trajectory point in the trajectory data; or, it can also be implemented by outlier detection, for example, by using statistical methods such as standard deviation and interquartile range or clustering methods such as density-based spatial clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) to detect and remove abnormal points in the trajectory data. This specification does not limit the specific denoising processing means.

[0052] Step S204: sampling the trajectory data.

[0053] In the case where the acquired trajectory data contains positioning information of multiple trajectory points, a portion of trajectory points may be selected to participate in the calculation of the above-mentioned trajectory features, so as to reduce the amount of data while maintaining the main features of the trajectory.

[0054] Specifically, the sampling method can be implemented by at least one of the following methods: time interval sampling: selecting trajectory points at fixed time intervals; distance interval sampling: selecting trajectory points at fixed geographical distances; key point sampling: selecting key points in the moving trajectory, such as turning points, acceleration points, etc.

[0055] Step S206, calculating corresponding trajectory features according to the trajectory data.

[0056] The trajectory features corresponding to the trajectory data after denoising and sampling can be further divided into the first type of trajectory features and the second type of trajectory features according to their specific calculation methods. Specifically, the first type of trajectory features are determined based on the positioning information of each trajectory point in the trajectory data, such as using the position information of each trajectory point in the positioning information as the position feature corresponding to the trajectory point, and using the time information of each trajectory point in the positioning information as the time feature corresponding to the trajectory point; the second type of trajectory features are determined by the positioning information between adjacent trajectory points, such as speed features, slope features, curvature features, trajectory quality features and other additional features that cannot be directly provided by the positioning information.

[0057] The following is a detailed description of the process of determining the above additional features:

[0058] For the speed feature, it can be calculated by dividing the distance between two adjacent trajectory points i and i+1 on the moving trajectory by the time difference, where i can correspond to the sequence of each trajectory point. Taking trajectory point i as an example, assuming that the trajectory data contains the timestamp t i and the corresponding position coordinates (x i ,y i ), then the speed v can be calculated by the following formula:

[0059]

[0060] Among them, (x i+1 ,y i+1 ) and (x i ,y i ) are the positioning coordinates of two adjacent trajectory points, i.e., the position information in the above positioning information; t i+1 and t i is the corresponding timestamp, that is, the time information in the above positioning information. Of course, although the above speed v is determined by the above formula to be the average moving speed of the target object characterized by the user between two adjacent trajectory points, when the time interval between two adjacent trajectory points is less than the preset interval length, the above speed v can be approximated as the speed feature of the previous trajectory point or the next trajectory point. In particular, when the trajectory point is the first trajectory point in the trajectory data, and the above speed v is used as the speed feature of the next trajectory point, the above first trajectory point can be set to the same speed feature as the second trajectory point; when the trajectory point is the last trajectory point in the trajectory data, and the above speed v is used as the speed feature of the previous trajectory point, the above last trajectory point can be set to the same speed feature as the second to last trajectory point.

[0061] For the slope feature, the elevation information in the trajectory data can be introduced to calculate the slope. Assume that the trajectory data contains elevation information h i and horizontal distance d i , then the slope θ between two adjacent trajectory points can be calculated by the following formula:

[0062]

[0063] The above slope θ can be used as a slope feature to indicate the inclination of the road or ground between two adjacent trajectory points. i+1 and h i is the elevation information between two adjacent trajectory points, which can be obtained from the position information in the positioning information; d i+1 and d iis the horizontal distance between two adjacent trajectory points. The horizontal distance can be calculated according to a spherical geometry formula such as the Haversine formula or the Vincenty formula, and this specification does not limit this.

[0064] For the curvature feature, it can be calculated by the relative positions between three adjacent trajectory points on the moving trajectory. Assume that the position information contained in the trajectory data is expressed as coordinates (x i ,y i ), the curvature K can be calculated by the following formula:

[0065]

[0066] Here, K can be used as a curvature feature to represent the curvature of the local moving trajectory formed by the above three trajectory points i-1, i and i+1. i+1 ,y i+1 ), (x i ,y i ), and (x i-1 ,y i-1 ) can be the coordinates of three consecutive adjacent trajectory points.

[0067] The trajectory quality characteristics can be calculated by evaluating the integrity, accuracy, clarity and continuity of the trajectory data. Specifically, the calculation of trajectory quality characteristics can be determined by establishing a comprehensive evaluation index system, in which a number of basic indicators including time interval quality, point loss rate quality, and track smoothness quality can be detected and recorded. By extracting the time information and location information from the above trajectory data, and using this information to calculate the above basic indicators, and then determining the weighted coefficients corresponding to each quality indicator based on the hierarchical analysis method, the comprehensive quality evaluation index of the above trajectory data is calculated by weighted summation.

[0068] Those skilled in the art can understand that the above-mentioned steps S202 and S204 are not necessary prerequisite steps for step S206. In fact, when the trajectory data has high accuracy and a small amount of data, the acquired trajectory data can be directly used to calculate trajectory features, thereby increasing the acquisition speed of trajectory features and reducing the computing resources required for the overall process.

[0069] In addition, unlike the first type of trajectory features, the second type of trajectory features, as additional features calculated based on trajectory data, introduces trajectory features of multiple different dimensions to represent the movement trajectory, thereby reducing the negative impact of single-dimensional trajectory features on the movement trajectory.

[0070] Combine the following Figure 3 , the process of determining semantic features is further explained, such as Figure 3 As shown, the process includes the following steps:

[0071] Step S302: identifying a semantic scene according to the semantic data.

[0072] As mentioned above, the above semantic features can be determined based on semantic data alone or based on Figure 2 The trajectory features obtained by the method shown are determined by the above semantic data. The former means that the acquisition process of semantic features and trajectory features is independent of each other and can be processed in parallel to improve execution efficiency; the latter means that the semantic features will be jointly determined by data at different levels, avoiding the misleading of the semantic scene recognition process by a single aspect of data, and effectively improving the accuracy of the semantic features.

[0073] For semantic data, the road traffic information represented by it can be identified as a concrete semantic scene, so that the above-mentioned semantic features can be specifically expressed. The semantic scene can be used to describe any one or any combination of the road type of the road where the current moving trajectory is located, the road attributes of the road where the road is located, the lane attributes of the road where the road is located, the area information to which the road belongs, etc. The aforementioned road types may include but are not limited to elevated roads, highways, ramps, tunnels, mountain roads, roundabouts, bridges, pedestrian streets, squares, etc.; the aforementioned road attributes may include but are not limited to any one or any combination of main roads, secondary roads, express lanes, dedicated lanes, temporary roads, seasonal roads, tidal lanes, etc.; the aforementioned lane attributes may include but are not limited to any one or any combination of the number of lanes, the lane directions of each lane contained in the road, the lane sizes, etc.; the aforementioned area information may include but are not limited to any one or any combination of the natural landforms on both sides of the road, surrounding facilities, etc. The method of identifying semantic scenes based on semantic data specifically refers to: identifying the road scene corresponding to the above-mentioned moving trajectory as a semantic feature based on the above-mentioned semantic data, then the scheme for determining semantic features by combining the above-mentioned trajectory features and semantic data can actually be regarded as identifying the road scene corresponding to the above-mentioned moving trajectory based on the above-mentioned trajectory features and the above-mentioned semantic data as the above-mentioned semantic feature. In addition, this specification does not limit the implementation scheme of identifying road traffic information as a semantic scene, which can be achieved through methods such as convolutional neural networks (CNN) based on deep learning, rule-based methods, and knowledge graph-based methods. The following will introduce the method of combining the feature dictionary with the scene classification model, a machine learning model:

[0074] First, as a manually designed feature set, the feature dictionary can be used to describe the features of various aspects of the above trajectory data. At the same time, each feature has a unique code that can be mapped to an index. In short, each entry in the feature dictionary can contain the feature name of the corresponding feature, the feature type of trajectory feature or semantic feature, the feature code, and the corresponding feature description. Under this technical premise, it is assumed that the scenes that a certain brand of car needs to identify include: highways, urban roads, rural roads, and residential roads. In order to achieve this goal, the following features that are helpful for scene classification can be selected: speed features, slope features, curvature features, trajectory quality features determined by trajectory data, and road types determined based on road images taken by the car's road sensor, or the road type of the current road obtained by the car's built-in reference map. It is worth mentioning that the above trajectory quality features can be used to determine special road types such as tunnels where the trajectory feature acquisition capability is poor. Then, based on the pre-built feature dictionary, the above selected features are converted into structured feature vectors. For features that are not in the dictionary, a default value can be assigned or other special processing can be performed. Then, the obtained feature vectors are normalized to eliminate the dimensional influence between different features.

[0075] Then, you can choose a model that can output probability distribution, such as a logistic regression model, a random forest model, or a deep learning model based on the softmax activation function. In order to output a probability value for each category to indicate the possibility that the model believes the input data belongs to that category, the model classification capability is then trained using a pre-labeled historical scene dataset. Each sample contains a feature vector and a label, and after training, the model parameters are adjusted to obtain better performance. Finally, the corresponding feature vector is extracted from the new trajectory features, or the combination of trajectory features and semantic data, and the trained model is used for scene recognition. The probability values ​​between the roads corresponding to the mobile trajectory and highways, urban roads, rural roads, and residential roads are obtained, that is, the model's confidence in the output results.

[0076] Step S304: Using the identified semantic scene as a semantic feature.

[0077] The identified semantic scenes can be used as semantic features for subsequent road matching. Still taking the above scene classification model as an example, for multiple identified semantic scenes, the semantic scene with the highest confidence or multiple semantic scenes with the highest confidence can be used as semantic features according to the confidence level.

[0078] Step S106: determining a target path in a target map that matches the moving trajectory and the road traffic information according to the trajectory feature and the semantic feature.

[0079] The trajectory features and semantic features obtained above can be used in conjunction with the target map to determine the target path that the target object's moving trajectory actually passes through, wherein the target map may include a road network composed of multiple road sections, and the target path may be a continuous complex road composed of multiple road sections in the road network.

[0080] The process of determining the above target path can be divided into three steps. Figure 4 Explain the above three steps, such as Figure 4 As shown in the figure, the process of determining the target path includes the following steps:

[0081] Step S402: determining a plurality of candidate map sections matching the movement trajectory according to the trajectory feature.

[0082] First, the moving trajectory represented by the above trajectory data can be mapped to the above target map, and the mapped area can be used as the area to be searched. The area to be searched can be a rectangular area containing the starting point and the end point of the current moving trajectory after calculating the longitude and latitude of the two points, or it can be an area formed by a circle with the center point / end point of the current moving trajectory as the center and the length of the current moving trajectory as the radius. In short, its essence is to include every possible section related to the above moving trajectory.

[0083] Then, multiple road sections can be screened out from the area to be searched as candidate map sections. At this time, the above-mentioned candidate map sections can be determined based on the trajectory characteristics of the current trajectory and the distances and / or angles of each road section in the area to be searched. For example, for each road section, the distances from its starting point and end point to the center point of the current trajectory can be calculated, and a road section with a distance less than a preset road section distance threshold can be selected as an alternative map section; or the starting direction of the current moving trajectory, that is, the direction from the starting point to the end point, and the starting direction of each road section can be calculated, and the angle between the current moving trajectory direction and the direction of each road section can be calculated, and the road section with an angle less than a preset road section angle threshold can be selected as an alternative map section.

[0084] Step S404: determining a matching score between each alternative map segment and the moving trajectory based on the trajectory feature and the semantic feature, wherein the matching score is positively correlated with a first similarity between the moving trajectory and the corresponding alternative map segment and a second similarity between the road traffic information corresponding to the moving trajectory and the road traffic information in the corresponding alternative map segment.

[0085] After determining multiple candidate map sections in the search area, the matching score between each candidate map section and the moving trajectory can be evaluated by using the previously acquired trajectory features and corresponding semantic features. Figure 5The determination and calculation process of the above matching score is introduced, and the process may include:

[0086] Step S502: Calculate the first similarity and the second similarity.

[0087] In order to determine the above matching score, it is necessary to first calculate the first similarity between the moving trajectory and the corresponding alternative map section, and the second similarity between the road traffic information corresponding to the moving trajectory and the road traffic information in the corresponding alternative map section. Among them, the first similarity can be understood as the matching score for the trajectory feature, which can be obtained by comprehensively calculating the geometric distance and angle error between the moving trajectory corresponding to the trajectory feature and the alternative map section. For example, the closer the distance between the moving trajectory and the alternative map section and the smaller the angle difference, the higher the matching score, otherwise the first similarity is lower. The second similarity can be understood as the matching score for the semantic feature, which is obtained by comparing the semantic features recognized by the scene with the scene category marked by the alternative map section. For example, if the categories are the same, the second similarity is high. Of course, a penalty mechanism can also be introduced to reduce the second similarity when the categories are different. In addition to the comparison of scene categories, semantic features involving quantity such as the number of lanes can also be matched. For example, the closer the number of lanes between the two is, the higher the second similarity, otherwise the second similarity is lower.

[0088] In particular, different methods may be used to determine the above matching score for the positioning features, time features, speed features, slope features, curvature features, and trajectory quality features obtained above.

[0089] In one embodiment, when the above-mentioned trajectory features include at least one of the positioning features, speed features, slope features and curvature features, the feature difference between each of the above-mentioned alternative map sections and the above-mentioned moving trajectory for at least one of the above-mentioned features can be calculated, and then the above-mentioned matching score is determined based on the feature difference, wherein the feature difference is negatively correlated with the above-mentioned matching score. For example, each of the above-mentioned alternative map sections respectively records the vehicle prescribed driving speed of the section, then the speed represented by the speed feature of the above-mentioned moving trajectory corresponding to the section can be compared with the above-mentioned vehicle prescribed driving speed, and the speed difference between the two can be used as the feature difference. The matching score corresponding to the feature difference can be determined according to a preset proportional function. Obviously, the larger the feature difference, the lower the matching score corresponding to the above-mentioned speed feature. It can be understood that the above-mentioned time feature can be used to determine the speed feature, but the time feature itself does not need to participate in the calculation of the matching score.

[0090] In another embodiment, when the above-mentioned trajectory features include trajectory quality features, the score weight of the trajectory features corresponding to the above-mentioned trajectory quality features can be calculated, and the above-mentioned matching score can be determined according to the above-mentioned score weight, wherein the above-mentioned trajectory quality features are positively correlated with the above-mentioned score weight and the matching score, respectively. As mentioned above, the above-mentioned trajectory quality features can characterize the comprehensive quality evaluation index of the above-mentioned trajectory data, so the higher the comprehensive quality evaluation index corresponding to the trajectory quality features, the higher the score weight of the corresponding trajectory features, and the higher the matching score determined by the score weight, and vice versa, the lower the score weight of the corresponding trajectory features, and the lower the matching score determined by the score weight.

[0091] Step S504, calculating the weight of each feature in the similarity.

[0092] As mentioned above, the corresponding confidence can be obtained in the process of acquiring the above semantic features. In fact, the corresponding confidence can also be obtained in the process of acquiring trajectory features. For the confidence of different features, the matching score weights of different features can be obtained by normalization.

[0093] Step S506, calculating the matching score corresponding to each candidate map section.

[0094] Based on the matching score weights of different features obtained in the previous step, the first similarity and the second similarity can be multiplied by the matching score weights of the corresponding features, thereby obtaining a matching score with a weight. In short, the matching score as a whole shows a positive correlation with the first similarity and the second similarity, that is, the higher the first similarity or the second similarity, the higher the matching score, and vice versa, which makes the matching score actually correctly reflect the size relationship between the first similarity and the second similarity of any candidate map section and other candidate map sections.

[0095] It is understandable that the above weights can be directly set to preset default weight values ​​instead of being determined by the confidence of different features, thereby improving the calculation efficiency of the matching score at the expense of accuracy.

[0096] Step S406: determining the target path according to the matching score and the map data of the target map.

[0097] After determining the above matching score, the above matching score can be arranged from large to small, and then the first n candidate map sections with the highest matching scores can be selected, wherein n can be determined as any positive integer according to actual needs. In addition, all road networks can be determined from the map data of the above target map, and the state transition probability can be calculated according to the matching score of the candidate map section, which refers to the probability of transferring from one state (such as a road section) to another state (another road section). The conflicting candidate map sections can be converted when the paths are discontinuous in different candidate map sections, thereby improving the continuity of the target path. Finally, the above selected candidate map sections can be input as observation probabilities together with the corresponding state transition probabilities into advanced map matching algorithms of models such as Hidden Markov Model (HMM), Conditional Random Field (CRF), and intelligent optimization to calculate the target path consisting of at least multiple road sections.

[0098] In addition, the specification can also directly use the candidate map path with the largest matching score as the above target path. Since the implementation of this embodiment is simple and requires low calculation amount, it is suitable for scenes with low complexity of the target map.

[0099] Figure 6 is a schematic structural diagram of a device provided by an exemplary embodiment. Figure 6 At the hardware level, the device includes a processor 602, an internal bus 604, a network interface 606, a memory 608, and a non-volatile memory 610, and may also include hardware required for other functions. One or more embodiments of this specification may be implemented based on software, such as the processor 602 reading the corresponding computer program from the non-volatile memory 610 into the memory 608 and then running it. Of course, in addition to the software implementation, one or more embodiments of this specification do not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0100] Corresponding to the above-mentioned map matching method embodiment, this specification also provides an embodiment of a map matching device.

[0101] Please refer to Figure 7 ; The device comprises:

[0102] The data acquisition unit 702 is used to acquire trajectory data of a target object and its corresponding semantic data, wherein the trajectory data is used to represent the moving trajectory of the target object, and the semantic data is used to represent the road traffic information corresponding to the moving trajectory.

[0103] The feature determination unit 704 is used to determine the trajectory feature corresponding to the trajectory data and determine the semantic feature corresponding to the semantic data.

[0104] The path determination unit 706 is used to determine a target path in a target map that matches the moving trajectory and the road traffic information according to the trajectory feature and the semantic feature.

[0105] Optionally, the trajectory data includes positioning information of multiple trajectory points; the feature determination unit 704 is specifically used to:

[0106] Determine a first type of trajectory feature of each trajectory point according to the positioning information of each trajectory point in the trajectory data;

[0107] The second type of trajectory features of each trajectory point are determined according to the positioning information between adjacent trajectory points in the trajectory data, and the second type of trajectory features and the first type of trajectory features are used as the trajectory features.

[0108] Optionally, the first type of trajectory features includes at least any one of the following: positioning features, time features; and / or,

[0109] The second type of trajectory features includes at least any one of the following: speed features, slope features, curvature features, and trajectory quality features.

[0110] Optionally, the path determination unit 706 is specifically configured to:

[0111] Determining, according to the trajectory features, a plurality of candidate map sections that match the movement trajectory;

[0112] determining a matching score between each candidate map segment and the movement trajectory according to the trajectory feature and the semantic feature, wherein the matching score is positively correlated with a first similarity between the movement trajectory and the corresponding candidate map segment and a second similarity between the road traffic information corresponding to the movement trajectory and the road traffic information in the corresponding candidate map segment;

[0113] The target path is determined according to the matching score and the map data of the target map.

[0114] Optionally, the path determination unit 706 is specifically configured to:

[0115] In a case where the trajectory feature includes at least any one of a positioning feature, a speed feature, a slope feature, and a curvature feature, calculating a feature difference between each candidate map segment and the moving trajectory for the at least any one of the features, and determining the matching score according to the feature difference, wherein the feature difference is negatively correlated with the matching score;

[0116] In the case where the trajectory feature includes a trajectory quality feature, a score weight of the trajectory feature corresponding to the trajectory quality feature is calculated, and the matching score is determined according to the score weight, and the trajectory quality feature is positively correlated with the score weight and the matching score, respectively.

[0117] Optionally, the semantic feature is determined according to semantic data, or the semantic feature is determined according to the trajectory feature and the semantic data; and / or,

[0118] The semantic data is acquired through a road sensor of the target object; and / or, the semantic data is acquired through a reference map.

[0119] Optionally, the feature determination unit 704 is specifically configured to:

[0120] A road scene corresponding to the moving trajectory is identified according to the trajectory feature and the semantic data, and the road scene is used as the semantic feature.

[0121] Based on the same concept as the above method, this specification also provides an electronic device, including: a processor; a memory for storing processor executable instructions; wherein the processor implements the steps of the method described in any of the above embodiments by running the executable instructions.

[0122] Based on the same concept as the above method, this specification also provides a computer-readable storage medium on which computer instructions are stored. When the instructions are executed by a processor, the steps of the method described in any of the above embodiments are implemented.

[0123] Based on the same concept as the above method, this specification also provides a computer program product, including a computer program / instruction, which implements the steps of the method described in any of the above embodiments when executed by a processor.

[0124] Although the present invention includes many specific implementation details, these should not be interpreted as limiting the scope of any invention or the scope of protection claimed, but are mainly used to describe the features of the specific embodiments of specific inventions. Certain features described in multiple embodiments of the present invention may also be implemented in combination in a single embodiment. On the other hand, the various features described in a single embodiment may also be implemented separately in multiple embodiments or in any suitable sub-combination. In addition, although features may work in certain combinations as described above and even initially claimed as such, one or more features from the claimed combination may be removed from the combination in some cases, and the claimed combination may point to a sub-combination or a variation of a sub-combination.

[0125] Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that these operations be performed in the particular order shown or performed sequentially, or requiring that all illustrated operations be performed to achieve the desired results. In some cases, multitasking and parallel processing may be advantageous. In addition, the separation of various system modules and components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product, or packaged into multiple software products.

[0126] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In some implementations, multitasking and parallel processing may be advantageous.

[0127] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A map matching method based on semantic features, characterized in that: The method comprises: Acquire trajectory data of a target object and its corresponding semantic data, wherein the trajectory data is used to characterize a moving trajectory of the target object, and the semantic data is used to characterize road traffic information corresponding to the moving trajectory; Determining a trajectory feature corresponding to the trajectory data, and determining a semantic feature corresponding to the semantic data; A target path in a target map that matches the moving trajectory and the road traffic information is determined according to the trajectory feature and the semantic feature.

2. The method according to claim 1, characterized in that The trajectory data includes positioning information of a plurality of trajectory points; and determining the trajectory features corresponding to the trajectory data includes: Determine a first type of trajectory feature of each trajectory point according to the positioning information of each trajectory point in the trajectory data; The second type of trajectory features of each trajectory point are determined according to the positioning information between adjacent trajectory points in the trajectory data, and the second type of trajectory features and the first type of trajectory features are used as the trajectory features.

3. The method according to claim 2, characterized in that The first type of trajectory features includes at least any one of the following: positioning features, time features; and / or, The second type of trajectory features includes at least any one of the following: speed features, slope features, curvature features, and trajectory quality features.

4. The method according to claim 1, characterized in that: The determining, according to the trajectory feature and the semantic feature, a target path in the target map that matches the movement trajectory and the road traffic information includes: Determining, according to the trajectory features, a plurality of candidate map sections that match the movement trajectory; determining a matching score between each candidate map segment and the movement trajectory according to the trajectory feature and the semantic feature, wherein the matching score is positively correlated with a first similarity between the movement trajectory and the corresponding candidate map segment and a second similarity between the road traffic information corresponding to the movement trajectory and the road traffic information in the corresponding candidate map segment; The target path is determined according to the matching score and the map data of the target map.

5. The method according to claim 4, characterized in that The determining a matching score between each candidate map segment and the moving trajectory according to the trajectory feature and the semantic feature includes: In a case where the trajectory feature includes at least any one of a positioning feature, a speed feature, a slope feature, and a curvature feature, calculating a feature difference between each candidate map segment and the moving trajectory for the at least any one of the features, and determining the matching score according to the feature difference, wherein the feature difference is negatively correlated with the matching score; In the case where the trajectory feature includes a trajectory quality feature, a score weight of the trajectory feature corresponding to the trajectory quality feature is calculated, and the matching score is determined according to the score weight, and the trajectory quality feature is positively correlated with the score weight and the matching score, respectively.

6. The method according to claim 1, characterized in that The semantic feature is determined according to the semantic data, or the semantic feature is determined according to the trajectory feature and the semantic data; and / or, The semantic data is acquired through a road sensor of the target object; and / or, the semantic data is acquired through a reference map.

7. The method according to claim 6, characterized in that Determining the semantic feature according to the trajectory feature and the semantic data includes: A road scene corresponding to the moving trajectory is identified according to the trajectory feature and the semantic data, and the road scene is used as the semantic feature.

8. An electronic device, characterized in that: include: processor; A memory for storing processor-executable instructions; wherein the processor implements the steps of the method according to any one of claims 1 to 7 by executing the executable instructions.

9. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product, characterized in that The method comprises a computer program / instruction, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 7.

Images