Vehicle type identification method and device based on parking semantics, equipment and storage medium

Abstract

The invention discloses a vehicle type identification method and device based on parking semantics, equipment and a storage medium. Comprising the following steps: acquiring historical track data of a vehicle, determining a loading and unloading stop point sequence of the vehicle based on the historical track data, and recording point-of-interest information or region-of-interest information corresponding to stop points; obtaining a plurality of travels of the vehicle based on the loading and unloading stop point sequence; constructing a loading and unloading semantic word sequence of the vehicle according to the point-of-interest information or the region-of-interest information corresponding to the starting point and the ending point of each travel of the vehicle; and inputting the loading and unloading semantic word sequence into a pre-trained vehicle type recognition model, and outputting predicted vehicle type information. The invention provides a freight vehicle type judgment method based on parking semantics, and the method comprises the steps: representing the historical behavior of a vehicle as a freight semantic vector, and deducing the actual type of the vehicle through a machine learning model. And the prediction accuracy is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of logistics transportation, in particular to a vehicle type identification method and device based on parking semantics, equipment and storage medium. BACKGROUND

[0002] In actual freight scenarios, the tractor of a heavy semi-trailer truck is separated from the trailer, and the actual freight type depends on the type of the trailer connected. However, the existing freight data often has the problem of missing trailer type information, which brings many inconveniences to the application scenarios such as logistics map drawing, vehicle and freight matching, and industry chain analysis.

[0003] At present, some schemes are based on a vehicle type determination method for parking in an aggregation area. This method uses a label propagation algorithm to infer the vehicle type by analyzing the parking behavior of the vehicle in the aggregation area. However, this method has some limitations. First, the calculation overhead of the label propagation algorithm is large, which may lead to low efficiency in large-scale data processing, making it difficult to meet the requirements of real-time application scenarios. Second, this method only uses vehicle type distribution information, which has insufficient interpretability and verifiability, and cannot fully meet the requirements of vehicle transportation type identification accuracy and reliability in actual applications. SUMMARY

[0004] The embodiments of the present application provide a vehicle type identification method and device based on parking semantics, equipment and storage medium to at least solve the technical problem of difficulty in accurately identifying freight vehicle types in related technologies.

[0005] According to an aspect of an embodiment of the present application, a vehicle type identification method based on parking semantics is provided, comprising:

[0006] Obtaining historical trajectory data of a vehicle, determining a loading and unloading parking point sequence of the vehicle based on the historical trajectory data, and recording interest point information or interest area information corresponding to the parking points;

[0007] Based on the loading and unloading parking point sequence, obtaining a plurality of trips of the vehicle;

[0008] According to the interest point information or interest area information corresponding to the start and end points of each trip of the vehicle, a loading and unloading semantic word sequence of the vehicle is constructed;

[0009] The loading and unloading semantic word sequence is input into a pre-trained vehicle type identification model, and predicted vehicle type information is output.

[0010] In one embodiment, determining a loading and unloading parking point sequence of the vehicle based on the historical trajectory data, and recording interest point information or interest area information corresponding to the parking points; comprises:

[0011] acquiring a vehicle parking point based on the historical trajectory data;

[0012] generating an electronic fence based on the vehicle gathering area, obtaining a fence parking point based on a parking point within the fence, and recording a name of an interest area, a type of the interest area, and a type of goods of the fence parking point;

[0013] determining a loading and unloading parking point based on the parking point outside the fence, and recording a name of a point of interest, a type of the point of interest, and a type of goods corresponding to the loading and unloading parking point;

[0014] sorting the fence parking point and the loading and unloading parking point in time sequence to obtain a loading and unloading parking point sequence of the vehicle.

[0015] In an embodiment, based on the loading and unloading parking point sequence, a plurality of trips of the vehicle is obtained, including:

[0016] performing morphological feature recognition on the loading and unloading parking point sequence;

[0017] dividing the loading and unloading parking point sequence into a plurality of trips according to the recognized morphological features.

[0018] In an embodiment, a loading and unloading semantic word sequence of the vehicle is constructed according to the point of interest information or the interest area information corresponding to the start and end points of each trip of the vehicle, including:

[0019] obtaining a semantic text description of the start and end points of the trip according to the point of interest information or the interest area information corresponding to the start and end points of each trip of the vehicle, and performing word segmentation preprocessing on the semantic text description;

[0020] collecting all trips of the vehicle, and representing the all trips of the vehicle as a word vector sequence set of the vehicle, the word vector sequence set including a word vector sequence corresponding to each trip of the vehicle;

[0021] performing random sampling with replacement and sequence processing on the word vector sequence set to obtain the loading and unloading semantic word sequence of the vehicle.

[0022] In an embodiment, performing random sampling with replacement and sequence processing on the word vector sequence set includes:

[0023] performing random sampling with replacement on the word vector sequence set to obtain a sampled word vector sequence set;

[0024] setting a word vector sequence length threshold;

[0025] performing truncation or padding processing on the sampled word vector sequence based on the word vector sequence length threshold to obtain the sequence-processed loading and unloading semantic word sequence of the vehicle.

[0026] In an implementation, the semantic text description is subjected to word segmentation preprocessing, including:

[0027] The semantic text description is subjected to word segmentation processing based on a preset word segmentation model, and is decomposed into word or sub-word units;

[0028] Each word after word segmentation is mapped to a unique integer index in a vocabulary table;

[0029] The integer index is converted into a dense vector representation.

[0030] In an implementation, before the loading and unloading semantic word sequence is input into the pre-trained vehicle type recognition model, the method further includes:

[0031] A plurality of vehicle corresponding loading and unloading semantic word sequences and corresponding vehicle type labels are obtained to construct a training data set;

[0032] A neural network model is trained based on the training data set to obtain a trained vehicle type recognition model.

[0033] According to another aspect of the embodiments of the present application, a vehicle type recognition device based on parking semantics is provided, including:

[0034] A parking data analysis module is configured to obtain historical trajectory data of a vehicle, determine a loading and unloading parking point sequence of the vehicle based on the historical trajectory data, and record interest point information or interest area information corresponding to the parking points;

[0035] A trip calculation module is configured to obtain a plurality of trips of the vehicle based on the loading and unloading parking point sequence;

[0036] A parking semantic acquisition module is configured to construct a loading and unloading semantic word sequence of the vehicle according to interest point information or interest area information corresponding to the start and end points of each trip of the vehicle;

[0037] A vehicle type recognition module is configured to input the loading and unloading semantic word sequence into a pre-trained vehicle type recognition model, and output predicted vehicle type information.

[0038] According to another aspect of the embodiments of the present application, an electronic device is also provided, including a memory and a processor, the memory stores a computer program, and the processor is configured to execute the above-mentioned vehicle type recognition method based on parking semantics through the computer program.

[0039] According to another aspect of the embodiments of the present application, a computer readable storage medium is also provided, which stores a computer program, wherein the computer program is configured to execute the above-mentioned vehicle type recognition method based on parking semantics when running.

[0040] The technical solutions provided by the embodiments of the present application can have the following beneficial effects:

[0041] The scheme provided by the embodiment of the application is configured according to the interest point information or the interest area information corresponding to the parking point of the vehicle, and constructs a loading and unloading semantic word sequence of the vehicle; the loading and unloading semantic word sequence is input into a pre-trained vehicle type recognition model to output predicted vehicle type information. First, the loading and unloading semantic word sequence is constructed by using the interest point information or the interest area information of the vehicle parking point, which can fully mine the semantic features of the parking area, and more comprehensively reflects the actual transportation scene of the vehicle compared with the method of relying only on the vehicle type distribution in the prior art. Secondly, the loading and unloading semantic word sequence is input into the pre-trained vehicle type recognition model, which can efficiently and accurately output the predicted vehicle type information. The scheme can provide more accurate and reliable data support for application scenarios such as logistics map drawing, vehicle and cargo matching and industry chain analysis, and improve the overall logistics operation efficiency and management level. BRIEF DESCRIPTION OF DRAWINGS

[0042] The accompanying drawings, which are included to provide a further understanding of the application, constitute a part of this application, illustrate the illustrative embodiments of the application and the description thereof, and serve to explain the application without limiting the application. In the drawings:

[0043] Figure 1 is a flowchart of a vehicle type recognition method based on parking semantics according to an embodiment of the application;

[0044] Figure 2 is a flowchart of another vehicle type recognition method based on parking semantics according to an embodiment of the application;

[0045] Figure 3 is a schematic diagram of a vehicle type recognition device based on parking semantics according to an embodiment of the application;

[0046] Figure 4 is a structural schematic diagram of an optional electronic device according to an embodiment of the application. DETAILED DESCRIPTION

[0047] In order to enable those skilled in the art to better understand the scheme of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0048] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0049] The following will be described in detail in combination with the accompanying drawings Figures 1-2 The method for identifying the type of vehicle based on the parking semantics of the embodiments of the present application will be described in detail. As shown in the figure, the method mainly includes the following steps: Figure 1

[0050] S101 obtains historical trajectory data of the vehicle, determines the loading and unloading parking point sequence of the vehicle based on the historical trajectory data, and records the interest point information or interest area information corresponding to the parking point.

[0051] In one embodiment, the loading and unloading parking point sequence of the vehicle is determined based on the historical trajectory data, and the interest point information or interest area information corresponding to the parking point is recorded.

[0052] First, based on the historical trajectory data, the parking point of the vehicle is obtained.

[0053] Collect the historical trajectory data of the vehicle, which usually includes the position coordinates (such as GPS longitude and latitude) of the vehicle at different time points. Then, the trajectory data is preprocessed to remove noise points and abnormal data. Then, by analyzing the speed change or stay time in the trajectory, the parking point of the vehicle is identified. For example, when the speed of the vehicle is zero or the stay time exceeds a certain threshold, it can be determined that the position is a parking point. Finally, the identified parking points are summarized and arranged to form a data set of the vehicle parking points, providing a basis for subsequent analysis and application.

[0054] Further, an electronic fence is generated based on the vehicle aggregation area, the fence parking point is obtained based on the parking point in the fence, and the interest area name, interest area type and cargo type of the fence parking point are recorded.

[0055] ​By analyzing the historical trajectory data of the vehicle, vehicle gathering areas are identified, which can be loading and unloading points or other important stops. Then, these gathering areas are manually labeled to generate core information, including the fence range (i.e., the geographical boundary of the gathering area), the AOI (Area of Interest) name, the AOI type (such as loading and unloading area, storage area, etc.), and the possible cargo type. Then, the stop points of the vehicle are associated with these labeled fences, and the stop points are determined to be within a certain fence range by geographic coordinate calculation. For the stop points within the fence, the AOI name, AOI type, and possible cargo type information of the fence are passed to the stop point, thereby enriching the semantic information of the stop point and providing more detailed data support for subsequent analysis and application.

[0056] Further, based on the stop points outside the fence, loading and unloading stop points are determined, and the POI (Point of Interest) name, POI type, and cargo type corresponding to the loading and unloading stop points are recorded.

[0057] Since manual labeling has problems such as insufficient coverage and changes in the fence over time, relying solely on "whether the stop point is within the fence" cannot ensure the completeness of the loading and unloading points. Therefore, for stop points that do not hit the fence, information such as gathering area portrait, satellite images, and vehicle historical behavior is combined to dynamically calculate loading and unloading points as a supplement. Then, the algorithm-inferred loading and unloading points are associated with GIS (Geographic Information System) data to obtain the POI (Point of Interest) name and type information near the loading and unloading points. For POIs containing obvious keywords, the possible cargo type is further inferred through a keyword and cargo type mapping table, for example, "Express Wuhan Distribution Center" is associated with "express", "coal washing plant" is associated with "coal", and "daily chemical" is associated with "daily necessities".

[0058] Finally, based on the fence stop points and the loading and unloading stop points, the vehicle's loading and unloading stop point sequence is obtained in chronological order.

[0059] The fence stop points and the algorithm-inferred loading and unloading stop points are merged, grouped by vehicle ID, and sorted in chronological order for each vehicle's fence stop points and loading and unloading stop points, thereby forming a complete loading and unloading stop point sequence for each vehicle.

[0060] S102 obtains multiple trips of the vehicle based on the loading and unloading stop point sequence.

[0061] In one embodiment, based on the loading and unloading stop point sequence, multiple trips of the vehicle are obtained, including: performing morphological feature recognition on the loading and unloading stop point sequence; and dividing the loading and unloading stop point sequence into multiple trips according to the recognized morphological features.

[0062] The loading and unloading point positions are sequentially connected in chronological order to form a running track of the vehicle. The track is analyzed by using a shape method to calculate shape features of the trip, such as direction changes. When the running direction of the vehicle is deflected by a large angle, it is determined that the previous trip ends and the next trip starts, so as to divide the entire track into multiple independent trip segments.

[0063] In S103, a sequence of semantic words of the loading and unloading of the vehicle is constructed according to the information of the interest points or the information of the interest areas corresponding to the start and end points of each trip of the vehicle.

[0064] In an embodiment, first, valid trips are screened, and the transportation distance of the trip must be greater than a set threshold value, so as to exclude interference data caused by difficulty in judging the start and end points in short trips; second, at least one of the start point or the end point of the trip is located in the range of the labeled fence, so as to ensure that the trip is associated with the known important area; finally, the start point of the trip must have valid POI (point of interest) name or type information, so as to be able to obtain detailed semantic description about the start point. Only the trips that meet all the above conditions are regarded as valid trips and are retained for subsequent analysis and processing.

[0065] Further, semantic text descriptions of the start and end points of the trip are obtained according to the information of the interest points or the information of the interest areas corresponding to the start and end points of the vehicle, and the semantic text descriptions are preprocessed by word segmentation.

[0066] For each trip, a text description containing information such as enterprise name and cargo type is generated by using the AOI or POI information of the stop point, such as "a mechanical factory in Wuxi loading equipment" <to>Shanghai port unloading.

[0067] Further, the semantic text description is segmented and preprocessed. The semantic text description is segmented and processed based on a preset segmentation model to decompose it into word or sub-word units; each word after segmentation is mapped to a unique integer index in the vocabulary table; and the integer index is converted into a dense vector representation.

[0068] After segmentation is completed, a vocabulary table needs to be constructed. The vocabulary table is a collection of all non-repeating words, each word in the vocabulary table has a unique integer index. All words or sub-word units appearing in all segmented texts can be collected by traversing all segmented texts, and each word or sub-word unit is assigned a unique integer index. Each word after segmentation is mapped to an integer index in the vocabulary table. For each segmented text, find the index of each word in the vocabulary table, and convert the text into an integer sequence.

[0069] A suitable word embedding method is selected to convert the integer index into a dense vector. Common word embedding methods include Word2Vec, GloVe, BERT, etc. These methods can encode semantic information of words into vectors, so that words with similar semantics are closer in vector space.

[0070] The entire journey set of the vehicle is represented as a set of word vector sequences of the vehicle, and the set of word vector sequences includes a word vector sequence corresponding to each journey of the vehicle. After the above process, the entire journey set S of a vehicle can be represented as follows:

[0071] S={x1,…,x l}

[0072] where x i represents the word vector sequence of a specific journey of the vehicle, and its structure is as follows:

[0073]

[0074] where [TO] is a token that separates the start and end points of a journey, [SEP] is a token that separates different journeys, represents the word sequence of the start point semantic description, represents the word sequence of the end point semantic description:

[0075]

[0076] w in the sequence represents the dense vector representation of the word of the start or end point text description.

[0077] Further, the set of word vector sequences is randomly sampled with replacement and processed to obtain the loading and unloading semantic word sequence of the vehicle.

[0078] Randomly sample from set S with replacement to generate semantic word sequence of length f:

[0079] X = {(x1',..., x f ′)|x i ′∈S}

[0080] There are two reasons for this sampling method: 1. Vehicle type judgment is not sensitive to the order of the trip, adjusting the order will not significantly affect the judgment, and for long sequence text, more positive samples can be generated by multiple random sampling; 2. This method can enhance the text for short sequences. Thus, the original trip set and label data can be converted into word vector sequence and label data:

[0081]

[0082] Further, a word vector sequence length threshold is set; based on the word vector sequence length threshold, the sampled word vector sequence is truncated or padded to obtain a sequence-processed loading and unloading semantic word sequence of the vehicle.

[0083] Due to the large difference in activity of different freight vehicles and the distance between different trips, the length of the trip sequence in the same statistical time window is extremely uneven. Therefore, in order to ensure the calculation efficiency of the model, additional truncation and padding of the word vector sequence is required. Set the maximum sequence length threshold k, pad the sequence shorter than the length with 0, and truncate the sequence longer than the length. It should be particularly pointed out that the truncation point should be specified at the trip delimiter position to ensure the integrity of the trip, so it needs to be truncated first and then padded. This step ensures that all samples have the same dimension, and then generates a fixed size tensor output, which is convenient for subsequent batch processing.

[0084] Finally, a loading and unloading semantic word sequence of each vehicle v j is constructed The form is:

[0085] <Loading point 1 description> <to><unloading point 1 description> <sep><Load point 2 description> <to><unloading point 2 description> <sep> …… <sep><Load point n description> <to><Unloading point n description>.

[0086] S104 inputs the loading and unloading semantic word sequence into the pre-trained vehicle type recognition model and outputs predicted vehicle type information.

[0087] In an embodiment, before inputting the loading and unloading semantic word sequence into the pre-trained vehicle type recognition model, the method further comprises: obtaining loading and unloading semantic word sequences corresponding to multiple vehicles and corresponding vehicle type labels, constructing a training data set; training a neural network model based on the training data set to obtain a trained vehicle type recognition model.

[0088] To construct a neural network model capable of recognizing vehicle types, first, the loading and unloading semantic word sequences of multiple vehicles and their corresponding vehicle type labels are collected to form a training data set. These loading and unloading semantic word sequences can be word vector sequences after tokenization, index mapping, and word embedding processing, while the vehicle type labels are the model or category information of the vehicle. Then, a suitable neural network architecture is selected, such as a recurrent neural network (RNN), a long short-term memory network (LSTM), or a Transformer architecture, and the model parameters are initialized. During training, the data set is input into the model, the loss between the predicted label and the true label is calculated, and the model parameters are updated through backpropagation to minimize the loss function. After multiple iterations of training, a trained vehicle type recognition model is obtained, which can accurately predict the vehicle type according to the input loading and unloading semantic word sequence.

[0089] Further, the processed word sequence is input into the pre-trained vehicle type recognition model, which analyzes and calculates the input semantic information; finally, the model outputs the predicted vehicle type information, thereby realizing automatic recognition and classification of vehicle types.

[0090] To facilitate understanding of the method provided by the embodiments of the present application, the following will describe the embodiments of the present application with reference to the accompanying drawings. Figure 2 Further description.

[0091] As Figure 2 shown, the flowchart describes a process from fence annotation to model training or prediction. First, the fence is annotated, then the fence stop points are calculated, the AOI information of the fence is obtained for the stop points that hit the fence, and the surrounding effective POI information is obtained for the stop points that do not hit the fence. Then, the trajectory shape method is used to calculate the trip, generate the text representation of the trip, and convert the text into word vector representation, and finally the model is trained or predicted. This process involves multiple steps such as geographic information processing, trajectory analysis, text processing, and machine learning, aiming to understand and predict vehicle behavior by analyzing vehicle stop behavior and trajectory data.

[0092] The application realizes accurate identification and classification of vehicle parking behavior by combining fence marking, trajectory analysis, semantic processing and machine learning technology. This method not only improves the accuracy and automation of loading and unloading point identification, but also enhances the identification ability of unmarked area loading and unloading behavior through dynamic calculation and algorithm optimization. In addition, the use of word vectors and pre-trained models for text description and vehicle type prediction further improves the efficiency and accuracy of vehicle type identification. Overall, this application helps to optimize logistics management, improve transportation efficiency, and provides strong technical support for vehicle monitoring and scheduling, with significant social and economic benefits.

[0093] According to another aspect of the embodiments of the application, a parking semantic based vehicle type identification device for implementing the above parking semantic based vehicle type identification method is also provided. As shown in Figure 3 The device comprises:

[0094] The parking data analysis module 301 is configured to obtain historical trajectory data of the vehicle, determine a loading and unloading parking point sequence of the vehicle based on the historical trajectory data, and record interest point information or interest area information corresponding to the parking points;

[0095] The trip calculation module 302 is configured to obtain a plurality of trips of the vehicle based on the loading and unloading parking point sequence;

[0096] The parking semantic acquisition module 303 is configured to construct a loading and unloading semantic word sequence of the vehicle according to the interest point information or interest area information corresponding to the start and end points of each trip of the vehicle;

[0097] The vehicle type identification module 304 is configured to input the loading and unloading semantic word sequence into a pre-trained vehicle type identification model and output predicted vehicle type information.

[0098] It should be noted that the parking semantic based vehicle type identification device provided in the above embodiments is used to execute the parking semantic based vehicle type identification method, and only the division of the above functional modules is used as an example. In actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the parking semantic based vehicle type identification device provided in the above embodiments and the parking semantic based vehicle type identification method belong to the same concept, and the implementation process is described in detail in the method embodiments. Here, it is not repeated.

[0099] According to another aspect of the embodiments of the application, an electronic device corresponding to the parking semantic based vehicle type identification method provided in the above embodiments is also provided to execute the above parking semantic based vehicle type identification method.

[0100] Please refer to Figure 4 This illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 4 As shown, the electronic device includes: a processor 400, a memory 401, a bus 402, and a communication interface 403. The processor 400, the communication interface 403, and the memory 401 are connected via the bus 402. The memory 401 stores a computer program that can run on the processor 400. When the processor 400 runs the computer program, it executes the vehicle type recognition method based on parking semantics provided in any of the foregoing embodiments of this application.

[0101] The memory 401 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 403 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0102] Bus 402 can be an ISA bus, PCI bus, or EISA bus, etc. Buses can be divided into address buses, data buses, control buses, etc. Memory 401 is used to store programs. After receiving execution instructions, processor 400 executes the program. The vehicle model recognition method based on parking semantics disclosed in any of the aforementioned embodiments of this application can be applied to processor 400, or implemented by processor 400.

[0103] The processor 400 can be an integrated circuit chip with a processing capability of signals. In the implementation process, each step of the above method can be completed by the integrated logic circuit of hardware in the processor 400 or the instruction in the form of software. The processor 400 described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a ready programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. Each method, step and logic block diagram disclosed in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software modules in the code processor for execution. The software module can be located in a random access memory, a flash memory, a read only memory, a programmable read only memory or an electrically erasable programmable memory, a register, etc. The storage medium in the art. The storage medium is located in the memory 401, and the processor 400 reads the information in the memory 401, and combines the hardware to complete the steps of the above method.

[0104] The electronic device provided by the embodiments of the present application and the vehicle type identification method based on parking semantics provided by the embodiments of the present application have the same inventive concept and the same beneficial effects as the method adopted, run or implemented by the electronic device.

[0105] According to another aspect of the embodiments of the present application, a computer readable storage medium corresponding to the vehicle type identification method based on parking semantics provided by the preceding embodiments is also provided, and the computer readable storage medium has a computer program (i.e. program product) stored thereon. When the computer program is run by a processor, the vehicle type identification method based on parking semantics provided by any of the preceding embodiments is executed.

[0106] It should be noted that examples of the computer readable storage medium can also include, but are not limited to, a phase change memory (PRAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), other types of random access memory (RAM), a read only memory (ROM), an electrically erasable programmable read only memory (EEPROM), a flash memory or other optical, magnetic storage medium, which will not be described one by one here.

[0107] The computer readable storage medium provided by the above embodiments of the present application and the vehicle type identification method based on parking semantics provided by the embodiments of the present application have the same inventive concept and the same beneficial effects as the method adopted, run or implemented by the application program stored in the computer readable storage medium.

[0108] The technical features of the above embodiments can be combined in any manner. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described, however, as long as the combinations of the technical features do not contradict each other, they should be considered to be within the scope of the present disclosure.

[0109] The above embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the patent scope of the present application. It should be pointed out that, for ordinary skilled persons in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.< / to> < / sep> < / sep> < / to> < / sep> < / to> < / to>

Claims

1. A method for vehicle type recognition based on stop semantics, characterized in that, The method comprises the following steps: acquiring historical trajectory data of a vehicle, determining a sequence of loading and unloading stops of the vehicle based on the historical trajectory data, and recording interest point information or interest area information corresponding to the loading and unloading stops; based on the sequence of loading and unloading stops, obtaining multiple trips of the vehicle; constructing a sequence of loading and unloading semantic words of the vehicle according to the interest point information or interest area information corresponding to the start and end points of each trip of the vehicle; inputting the sequence of loading and unloading semantic words into a pre-trained vehicle type recognition model to output predicted vehicle type information.

2. The method of claim 1, wherein, determining a sequence of loading and unloading stops of a vehicle based on historical trajectory data and recording interest point information or interest area information corresponding to the loading and unloading stops; comprising: based on the historical trajectory data, obtaining vehicle stops; generating an electronic fence based on a vehicle aggregation area, obtaining fence stops based on stops within the fence, and recording the interest area name, interest area type, and cargo type of the fence stops; determining loading and unloading stops based on stops outside the fence, and recording the interest point name, interest point type, and cargo type corresponding to the loading and unloading stops; based on the fence stops and loading and unloading stops, sorting in time sequence to obtain a sequence of loading and unloading stops of the vehicle.

3. The method of claim 1, wherein, based on the sequence of loading and unloading stops, obtaining multiple trips of the vehicle, comprising: performing morphological feature recognition on the sequence of loading and unloading stops; according to the recognized morphological features, dividing the sequence of loading and unloading stops into multiple trips.

4. The method of claim 1, wherein, constructing a sequence of loading and unloading semantic words of the vehicle according to the interest point information or interest area information corresponding to the start and end points of each trip of the vehicle, comprising: obtaining semantic text descriptions of the start and end points of each trip of the vehicle according to the interest point information or interest area information corresponding to the start and end points of each trip of the vehicle, and performing word segmentation preprocessing on the semantic text descriptions; representing the entire trip set of the vehicle as a word vector sequence set of the vehicle, the word vector sequence set comprising a word vector sequence corresponding to each trip of the vehicle; performing random sampling with replacement and sequence processing on the word vector sequence set to obtain the sequence of loading and unloading semantic words of the vehicle.

5. The method of claim 4, wherein, performing random sampling with replacement and sequence processing on the word vector sequence set, comprising: performing random sampling with replacement on the word vector sequence set to obtain a sampled word vector sequence set; setting a word vector sequence length threshold; based on the word vector sequence length threshold, performing truncation or padding processing on the sampled word vector sequence to obtain the sequence of loading and unloading semantic words of the vehicle after sequence processing.

6. The method of claim 4, wherein, performing word segmentation preprocessing on the semantic text description, comprising: performing word segmentation processing on the semantic text description based on a pre-set word segmentation model to decompose it into single words or sub-word units; mapping each word after word segmentation to a unique integer index in a vocabulary table; converting the integer index to a dense vector representation.

7. The method of claim 1, wherein, Before inputting the sequence of loading and unloading semantic words into the pre-trained vehicle type recognition model, further comprising: obtaining the sequence of loading and unloading semantic words corresponding to multiple vehicles and the corresponding vehicle type labels to construct a training data set; training a neural network model based on the training data set to obtain a trained vehicle type recognition model.

8. A vehicle model recognition device based on a parking semantics, characterized by, ​ The parking data analysis module is configured to obtain historical trajectory data of the vehicle, determine a sequence of loading and unloading parking points of the vehicle based on the historical trajectory data, and record interest point information or interest area information corresponding to the parking points; The trip calculation module is configured to obtain a plurality of trips of the vehicle based on the sequence of loading and unloading parking points. The parking semantic acquisition module is configured to construct a sequence of loading and unloading semantic words of the vehicle according to the interest point information or interest area information corresponding to the start and end points of each trip of the vehicle. The vehicle type identification module is configured to input the sequence of loading and unloading semantic words into a pre-trained vehicle type identification model, and output predicted vehicle type information.

9. An electronic device, comprising: A processor and a memory storing program instructions, the processor being configured to execute the program instructions to perform the parking semantic based vehicle type identification method according to any one of claims 1 to 7.

10. A computer readable medium characterized by A computer readable instruction is stored thereon, and the computer readable instruction is executed by a processor to implement a parking semantic based vehicle type identification method according to any one of claims 1 to 7.

Images