Loading and unloading point identification model training method, loading and unloading point identification method and electronic equipment

Abstract

The invention provides a loading and unloading point recognition model training method, a loading and unloading point recognition method and electronic equipment. The method comprises the following steps: acquiring a plurality of training samples and category labels respectively corresponding to the plurality of training samples; inputting the plurality of training samples into an initial loading and unloading point identification model; and based on the category labels corresponding to the plurality of training samples, training the initial loading and unloading point identification model until a model convergence condition is satisfied, and obtaining a target loading and unloading point identification model. In the application, the training of the target loading and unloading point identification model depends on the fusion feature data of the first track point included in the training sample, and the fusion feature data of the first track point describes the first track point from multiple dimensions, so that a more comprehensive view angle is provided; the initial loading and unloading point recognition model can better recognize and learn features of different types of track points, the classification accuracy of the target loading and unloading point recognition model is improved, and then the loading and unloading point recognition accuracy of the target loading and unloading point recognition model is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a training method for a loading and unloading point recognition model, a loading and unloading point recognition method, and an electronic device. Background Art

[0002] In modern logistics and transportation, the data of online transportation orders are usually recorded and analyzed in detail, but a large amount of offline transportation order data has not been effectively captured and counted. The lack of offline transportation order data not only leads to incomplete logistics information, but also affects the accurate assessment of the driver's actual workload and transportation efficiency. Therefore, identifying the driver's loading and unloading points in offline transportation orders and obtaining offline transportation order data has become a key technical issue in improving the level of logistics management.

[0003] In the related art, the loading and unloading points of offline transport orders are usually identified by analyzing the trajectory data of offline transport orders. When the transportation vehicle stays at a certain location for more than a preset time threshold, the location is considered to be a loading or unloading point. However, this method is prone to recognition errors. For example, rest points that are not for loading and unloading (such as restaurants where drivers eat, temporary resting places) may be mistakenly identified as loading or unloading points, resulting in low accuracy in loading and unloading point recognition. Summary of the invention

[0004] The present application provides a training method for a loading and unloading point recognition model, a loading and unloading point recognition method and an electronic device to solve the problem of low loading and unloading point recognition accuracy in the prior art.

[0005] In a first aspect, the present application provides a method for training a loading and unloading point recognition model, the method comprising:

[0006] Acquire multiple training samples and category labels corresponding to the multiple training samples respectively; wherein each of the training samples includes fused feature data of a first trajectory point, the fused feature data of the first trajectory point includes travel feature data and business feature data of the first trajectory point, the travel feature data includes vehicle speed related data, position related data, spatial data, steering angle data and detour ratio data of the first trajectory point, and the business feature data includes loading data, unloading data, transport vehicle model data and driver data of the first trajectory point; the category label corresponding to each of the training samples is used to indicate the category of the first trajectory point, and the category label is a rest point label, a loading point label or an unloading point label;

[0007] Inputting the plurality of training samples into an initial loading and unloading point recognition model;

[0008] Based on the category labels corresponding to the multiple training samples, the initial loading and unloading point recognition model is trained until the model convergence condition is met, so as to obtain a target loading and unloading point recognition model; wherein the target loading and unloading point recognition model is used to predict the category of the second trajectory point, and the second trajectory point is any trajectory point corresponding to the target transport order.

[0009] In a possible design, before acquiring a plurality of training samples and the category labels respectively corresponding to the plurality of training samples, the method further includes:

[0010] Acquire order data of multiple historical transport orders; wherein the order data of each of the historical transport orders includes loading point location data, unloading point location data, transport vehicle model data and driver data of the historical transport order, and an initial trajectory point data set corresponding to the historical transport order; the initial trajectory point data set includes trajectory data corresponding to multiple third trajectory points respectively, and the trajectory data corresponding to each of the third trajectory points includes the location data of the third trajectory point, vehicle speed data and collection time point;

[0011] For each of the plurality of historical transport orders, based on the initial track point data set corresponding to the historical transport order, determine a first track point data set corresponding to the historical transport order; wherein the first track point data set includes track data corresponding to a plurality of first track points respectively;

[0012] For each first trajectory point of the plurality of first trajectory points, the training sample and the category label corresponding to the training sample are generated based on the trajectory data corresponding to the first trajectory point and the order information of the plurality of historical transportation orders.

[0013] In a possible design, for each historical transport order among the plurality of historical transport orders, determining a first track point data set corresponding to the historical transport order based on an initial track point data set corresponding to the historical transport order includes:

[0014] For each of the plurality of historical transport orders, determining an average vehicle speed of a first target trajectory point according to vehicle speed data of a fourth trajectory point, a collection time point of the fourth trajectory point, vehicle speed data of a fifth trajectory point, and a collection time point of the fifth trajectory point; wherein the first target trajectory point is any third trajectory point among the plurality of third trajectory points, the fourth trajectory point is a third trajectory point among the plurality of third trajectory points that is located before the first target trajectory point, and the fifth trajectory point is a third trajectory point among the plurality of third trajectory points that is located after the first target trajectory point;

[0015] In a case where an average vehicle speed of the first target trajectory point is zero, determining the first target trajectory point as the first trajectory point;

[0016] Based on the initial track point data set corresponding to the historical transport order and the first track point among the plurality of third track points, a first track point data set corresponding to the historical transport order is obtained.

[0017] In a possible design, for each of the plurality of first trajectory points, generating the training sample and the category label corresponding to the training sample based on the trajectory data corresponding to the first trajectory point and the order information of the plurality of historical transport orders includes:

[0018] For each of the plurality of first trajectory points, the training sample is generated based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transport order, as well as the loading point location data, unloading point location data, transport vehicle model data and driver data of the plurality of historical transport orders; wherein the target historical transport order is the historical transport order to which the first trajectory point belongs among the plurality of historical transport orders;

[0019] Based on the position data and collection time point of the first trajectory point, and the loading point position data and unloading point position data of the target historical transportation order, a category label corresponding to the training sample is generated.

[0020] In a possible design, generating the training sample based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transport order, as well as the loading point location data, unloading point location data, transport vehicle model data, and driver data of the multiple historical transport orders includes:

[0021] Generate travel feature data of the first trajectory point based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transportation order;

[0022] Generate business feature data of the first trajectory point based on the location data of the first trajectory point, and the loading point location data, unloading point location data, transportation vehicle model data, and driver data of the plurality of historical transportation orders;

[0023] fusing the travel feature data of the first trajectory point and the business feature data of the first trajectory point to obtain fused feature data of the first trajectory point;

[0024] The training sample is generated based on the fused feature data of the first trajectory point.

[0025] In a possible design, the generating of the category label corresponding to the training sample based on the location data and the collection time point of the first trajectory point, and the loading point location data and the unloading point location data of the target historical transport order includes:

[0026] determining a first distance based on the location data of the first trajectory point and the location data of the loading point of the target historical transport order;

[0027] determining a second distance based on the location data of the first trajectory point and the location data of the unloading point of the target historical transport order;

[0028] When the first distance is less than a first preset distance and the collection time point of the first trajectory point is within a first preset time period, the category label corresponding to the training sample is determined to be a loading point label; wherein the first preset distance is the radius of the loading area corresponding to the loading point of the historical transport order, and the first preset time period is the time period between a first moment when the vehicle first enters the loading area and a second moment when the vehicle first leaves the loading area;

[0029] When the second distance is less than the second preset distance and the collection time point of the first trajectory point is within the second preset time period, the category label corresponding to the training sample is determined to be the unloading point label; wherein the second preset distance is the radius of the unloading area corresponding to the unloading point of the historical transport order, and the second preset time period is the time period between the third moment when the transport vehicle first enters the unloading area and the fourth moment when the transport vehicle first leaves the unloading area;

[0030] When the first preset condition is met, determining that the category label corresponding to the training sample is a rest point label;

[0031] The first preset condition includes at least one of the following:

[0032] The first distance is not less than the first preset distance, and the second distance is not less than the second preset distance;

[0033] The first distance is smaller than the first preset distance, and the collection time point of the first trajectory point is outside the first preset time period;

[0034] The second distance is smaller than the second preset distance, and the collection time point of the first trajectory point is outside the second preset time period.

[0035] In a possible design, the model convergence condition is that the loss function of the initial loading and unloading point identification model in consecutive target rounds is less than the target loss function; or,

[0036] The change in the loss function of the initial loading and unloading point identification model in the continuous target number of rounds is less than the target loss function change.

[0037] In a second aspect, the present application provides a method for identifying a loading and unloading point, the method comprising:

[0038] Acquire a second trajectory point data set corresponding to the target transport order; wherein the second trajectory point data set includes trajectory data corresponding to a plurality of second trajectory points respectively;

[0039] For each of the multiple second trajectory points, the fused feature data of the second trajectory point is input into the target loading and unloading point recognition model as described in the first aspect or various possible designs of the first aspect above, to obtain a prediction result output by the target loading and unloading point recognition model; wherein the prediction result is used to indicate the category of the second trajectory point.

[0040] In a third aspect, the present application provides a training device for a loading and unloading point recognition model, the training device comprising:

[0041] memory and at least one processor;

[0042] The memory stores computer-executable instructions;

[0043] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the method described in the first aspect or various possible designs of the first aspect.

[0044] In a fourth aspect, the present application provides a loading and unloading point identification device, the device comprising:

[0045] memory and at least one processor;

[0046] The memory stores computer-executable instructions;

[0047] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the method described in the second aspect above.

[0048] In a fifth aspect, the present application provides an electronic device, comprising: a memory and at least one processor;

[0049] The memory stores computer-executable instructions;

[0050] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method described in the first aspect, various possible designs of the first aspect, or the second aspect.

[0051] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer execution instructions are stored. When the computer execution instructions are executed, the method described in the first aspect, various possible designs of the first aspect, or the second aspect is implemented.

[0052] The training method of the loading and unloading point recognition model provided in the embodiment of the present application firstly obtains a plurality of training samples and their corresponding category labels, wherein each training sample comprises fused feature data of a first trajectory point, the fused feature data comprises travel feature data and business feature data of the first trajectory point, the travel feature data comprises vehicle speed related data, position related data, spatial data, steering angle data and detour ratio data of the first trajectory point, the business feature data comprises loading data, unloading data, transport vehicle model data and driver data of the first trajectory point; the category label corresponding to each training sample is used to indicate the category of the first trajectory point, and the category label is a rest point label, a loading point label or an unloading point label; secondly, the plurality of training samples are input into the initial loading and unloading point recognition model; finally, the initial loading and unloading point recognition model is trained based on the category labels respectively corresponding to the plurality of training samples until the model convergence condition is met, thereby obtaining the target loading and unloading point recognition model. In the present application, the training of the target loading and unloading point recognition model relies on the fused feature data of the first trajectory point included in the training samples. The fused feature data of the first trajectory point describes the first trajectory point from multiple dimensions by integrating the travel feature data and the business feature data of the first trajectory point, providing a more comprehensive perspective, which helps the initial loading and unloading point recognition model to better identify and learn the features of trajectory points of different categories, thereby improving the classification accuracy of the target loading and unloading point recognition model, and thus effectively improving the recognition accuracy of the loading and unloading points when the target loading and unloading point recognition model is used to identify the loading and unloading points in actual applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A flowchart of a method for training a loading and unloading point recognition model provided in an embodiment of the present application;

[0054] Figure 2 A schematic diagram of the position relationship of a plurality of third trajectory points provided in an embodiment of the present application;

[0055] Figure 3 A schematic diagram of a flow chart of a method for identifying a loading and unloading point provided in an embodiment of the present application;

[0056] Figure 4 A schematic diagram of the structure of a training device for a loading and unloading point recognition model provided in an embodiment of the present application;

[0057] Figure 5 A schematic diagram of the structure of a loading and unloading point identification device provided in an embodiment of the present application;

[0058] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used in the specification of the application are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification, claims and drawings of this application are intended to cover non-exclusive inclusions.

[0061] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase "embodiments" in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0062] The term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists, A and B can exist at the same time, and B exists. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0063] In addition, the terms "first", "second", etc. in the specification and claims of this application or the above-mentioned drawings are used to distinguish different objects rather than to describe a specific order, and may explicitly or implicitly include one or more of the features.

[0064] In the description of this application, unless otherwise specified, "multiple" and "at least two" mean more than two (including two). Similarly, "multiple groups" and "at least two groups" mean more than two groups (including two).

[0065] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "connected" and "connection" should be understood in a broad sense. For example, "connected" or "connection" can refer to not only physical connection, but also electrical connection or signal connection. For example, it can be directly connected, that is, physically connected, or indirectly connected through at least one intermediate element, as long as the circuit is connected, or it can be the internal connection of two elements; signal connection can refer to signal connection through a circuit or through a media medium, such as radio waves. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0066] In order to make those skilled in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the accompanying drawings. It should be noted that different technical features in the present application can be combined with each other without conflict.

[0067] First, the terms involved in one or more embodiments of this specification are explained.

[0068] A stop point refers to the point in the trajectory data of a transport vehicle that corresponds to a certain location when the vehicle's stop time at that location reaches a certain threshold. The speed of the transport vehicle at the stop point is zero. A stop point can be a loading point, an unloading point, or a stop point.

[0069] The loading point refers to the trajectory point corresponding to a location in the trajectory data of the transport vehicle when the residence time of the transport vehicle at a certain location reaches a certain threshold and the cargo loading operation is performed at the location.

[0070] The unloading point refers to the trajectory point corresponding to a location in the trajectory data of the transport vehicle when the transport vehicle's residence time at a certain location reaches a certain threshold and the cargo unloading operation is performed at the location.

[0071] A rest point refers to a trajectory point corresponding to a location in the trajectory data of a transport vehicle when the stay time of the transport vehicle at a certain location reaches a certain threshold and no cargo loading or unloading operations are performed at the location.

[0072] Secondly, a training method for a loading and unloading point recognition model provided in an embodiment of the present application is explained.

[0073] Figure 1 A flow chart of a training method for a loading and unloading point recognition model provided in an embodiment of the present application. Figure 1 As shown, the training method of the loading and unloading point recognition model provided in the embodiment of the present application specifically includes S101 to S103, and S101 to S103 are described in detail below.

[0074] It should be noted that the execution entity of the training method of the loading and unloading point recognition model provided in the embodiment of the present application can be a controller.

[0075] S101. A controller obtains a plurality of training samples and category labels corresponding to the plurality of training samples.

[0076] Each training sample includes fused feature data of a first trajectory point, and the fused feature data of the first trajectory point includes travel feature data and business feature data of the first trajectory point.

[0077] It should be noted that the trip characteristic data includes vehicle speed-related data, position-related data, spatial data, steering angle data, and detour ratio data at the first trajectory point. The trip characteristic data can fully describe the dynamic behavior of the transport vehicle at the first trajectory point. The trip characteristic data provides multiple dimensions of input for the initial loading and unloading point recognition model, which helps to accurately identify and classify the types of stop points, such as loading points, unloading points, and rest points.

[0078] It should be noted that the business feature data includes the loading data, unloading data, transport model data and driver data of the first trajectory point. The business feature data can fully describe the business activities and environmental characteristics of the first trajectory point, and can support the initial loading and unloading point recognition model to accurately classify the first trajectory point.

[0079] The category label corresponding to each training sample is used to indicate the category of the first trajectory point, and the category label is a rest point label, a loading point label, or an unloading point label.

[0080] It should be noted that the categories of the first track points include rest points, loading points and unloading points.

[0081] When the category label corresponding to the training sample is a rest point label, the category of the first track point in the training sample is a rest point.

[0082] When the category label corresponding to the training sample is the loading point label, the category of the first track point in the training sample is the loading point.

[0083] When the category label corresponding to the training sample is the unloading point label, the category of the first trajectory point in the training sample is the unloading point.

[0084] S102: The controller inputs a plurality of training samples into an initial loading and unloading point recognition model.

[0085] Among them, after the controller obtains multiple training samples and the category labels corresponding to the multiple training samples, the controller inputs the obtained training samples and the category labels corresponding to the training samples into the initial loading and unloading point recognition model through data transmission or API interface.

[0086] It should be noted that the initial loading and unloading point recognition model is a machine learning model for identifying loading and unloading points. The initial loading and unloading point recognition model learns from training samples and understands the relationship between different features and loading and unloading point categories, thereby realizing the recognition of the second trajectory point of unknown category.

[0087] The initial loading and unloading point identification model may be a LightGBM model (Light Gradient Boosting Machine, LightGBM model) or a deep neural network (Deep Neural Network, DNN), which is not specifically limited in this embodiment.

[0088] For example, the initial loading and unloading point recognition model is the LightGBM model.

[0089] In this embodiment, by inputting multiple training samples into the initial loading and unloading point recognition model, the initial loading and unloading point recognition model can learn the relationship between richer features and loading and unloading point categories, thereby improving the recognition accuracy of the trained target loading and unloading point recognition model.

[0090] S103: The controller trains the initial loading and unloading point recognition model based on the category labels corresponding to the multiple training samples, until the model convergence condition is met, and obtains the target loading and unloading point recognition model.

[0091] Among them, the target loading and unloading point recognition model is used to predict the category of the second trajectory point, and the second trajectory point is any trajectory point corresponding to the target transportation order.

[0092] It should be noted that the controller iteratively trains the initial loading and unloading point recognition model based on multiple training samples and the category labels corresponding to the multiple training samples until the initial loading and unloading point recognition model meets the model convergence conditions, and then the trained target loading and unloading point recognition model can be obtained.

[0093] The above-mentioned controller iteratively trains the initial loading and unloading point recognition model based on multiple training samples and the category labels corresponding to the multiple training samples until the initial loading and unloading point recognition model meets the model convergence conditions. The method for obtaining the trained target loading and unloading point recognition model is the existing model training method, which will not be repeated in this embodiment.

[0094] In an embodiment of the present application, first, a plurality of training samples and their corresponding category labels are obtained, wherein each training sample includes fused feature data of a first trajectory point, the fused feature data includes travel feature data and business feature data of the first trajectory point, the travel feature data includes vehicle speed-related data, position-related data, spatial data, steering angle data and detour ratio data of the first trajectory point, and the business feature data includes loading data, unloading data, transport vehicle model data and driver data of the first trajectory point; the category label corresponding to each training sample is used to indicate the category of the first trajectory point, and the category label is a rest point label, a loading point label or an unloading point label; secondly, the plurality of training samples are input into an initial loading and unloading point recognition model; finally, the initial loading and unloading point recognition model is trained based on the category labels corresponding to the plurality of training samples until the model convergence conditions are met, thereby obtaining a target loading and unloading point recognition model. In the present application, the training of the target loading and unloading point recognition model relies on the fused feature data of the first trajectory point included in the training samples. The fused feature data of the first trajectory point describes the first trajectory point from multiple dimensions by integrating the travel feature data and the business feature data of the first trajectory point, providing a more comprehensive perspective, which helps the initial loading and unloading point recognition model to better identify and learn the features of trajectory points of different categories, thereby improving the classification accuracy of the target loading and unloading point recognition model, and thus effectively improving the recognition accuracy of the loading and unloading points when the target loading and unloading point recognition model is used to identify the loading and unloading points in actual applications.

[0095] In the above embodiment, the controller needs to obtain multiple training samples and the category labels corresponding to the multiple training samples. Next, the specific process of the controller obtaining multiple training samples and the category labels corresponding to the multiple training samples is described in detail.

[0096] In a possible embodiment, before the controller obtains multiple training samples and the category labels corresponding to the multiple training samples in S101, the training method of the loading and unloading point recognition model also includes steps a1 to a3, and steps a1 to a3 are described in detail below.

[0097] Step a1: The controller obtains order data of multiple historical transport orders.

[0098] Among them, historical transportation orders are online transportation orders that were successfully completed within the target time period.

[0099] It should be noted that the target time period can be set by the model management personnel, and this embodiment does not specifically limit this.

[0100] For example, the target time period is the past year, or the target time period is the past six months.

[0101] Among them, the order data of each historical transport order includes the loading point location data, unloading point location data, transport vehicle model data and driver data of the historical transport order, as well as the initial trajectory point data set corresponding to the historical transport order.

[0102] It should be noted that the loading point location data includes the longitude and latitude of the loading point location of the historical transport order.

[0103] The unloading point location data includes the longitude and latitude of the unloading point locations of the historical transportation orders.

[0104] Transport vehicle type data includes the type and length of transport vehicles used to complete historical transport orders.

[0105] For example, the transport vehicle type data of historical transport orders include a 1.8-meter van, a 1.8-meter refrigerated truck, a 2.7-meter van, a 2.7-meter refrigerated truck, a 4.2-meter van, a 4.2-meter refrigerated truck, and so on.

[0106] Driver data includes the name, driving experience, and ID number of the driver who completed historical transport orders.

[0107] It should be noted that the initial trajectory point data set corresponding to the historical transport order includes trajectory data corresponding to multiple third trajectory points corresponding to the historical transport order. The trajectory data corresponding to each third trajectory point includes the position data, vehicle speed data and collection time point of the third trajectory point.

[0108] The trajectory data corresponding to the plurality of third trajectory points in the initial trajectory point data set are arranged in chronological order according to the acquisition time points of the plurality of third trajectory points.

[0109] The position data of the third track point includes the longitude and latitude of the third track point.

[0110] The vehicle speed data at the third trajectory point includes an instantaneous vehicle speed of the transport vehicle at the third trajectory point.

[0111] The collection time point of the third trajectory point is the time point when the third trajectory point is collected.

[0112] For example, the order data of a historical transport order is: loading point longitude: 121.4737; loading point latitude: 31.2304; unloading point longitude: 114.3050; unloading point latitude: 30.5931; transport vehicle type data: 2.7-meter heavy truck; driver name: Zhang San; driver driving experience: 10 years; driver ID number: 123456789012345678; the initial track point data set includes track point 1, track point 2 and track point 3; track point 1 longitude: 121.4700; track point 1 latitude: 31.2300; track point 1 instantaneous vehicle speed: 60km / h; track point 1 collection time: 2024-08-01 08:00:00; longitude of track point 2: 121.4710; latitude of track point 2: 31.2310; instantaneous vehicle speed of track point 2: 55km / h; collection time of track point 2: 2024-08-01 08:05:00; longitude of track point 3: 121.4720; latitude of track point 3: 31.2320; instantaneous vehicle speed of track point 3: 50km / h; collection time of track point 3: 2024-08-01 08:10:00.

[0113] In this embodiment, by acquiring order data of multiple historical transport orders, real scene information required for training the initial loading and unloading point recognition model is provided, which provides a basis for subsequent training sample generation, so that the training of the initial loading and unloading point recognition model can be based on actual transport data, thereby improving the practicality and accuracy of the target loading and unloading point recognition model.

[0114] Step a2: for each historical transport order among the multiple historical transport orders, the controller determines a first track point data set corresponding to the historical transport order based on the initial track point data set corresponding to the historical transport order.

[0115] The first trajectory point data set includes trajectory data corresponding to a plurality of first trajectory points.

[0116] The trajectory data corresponding to the multiple first trajectory points in the first trajectory point data set are arranged in chronological order according to the collection time points of the multiple first trajectory points.

[0117] It should be noted that the first trajectory point is the third trajectory point in the initial trajectory point data set where the average vehicle speed of the transport vehicle is 0. The first trajectory point indicates that the transport vehicle is in a stop state at the trajectory point, that is, the first trajectory point may be a loading point, an unloading point, or a rest point.

[0118] In this embodiment, for each historical transport order, among all the third trajectory points of the historical transport order, a third trajectory point where the average vehicle speed of the transport vehicle is zero is selected as the first trajectory point, and a first trajectory point data set corresponding to the historical transport order is generated based on the selected first trajectory point. By determining the first trajectory point data set corresponding to the historical transport order, it can be ensured that the selected first trajectory point is a stop point of the transport vehicle, thereby improving the data quality of the training samples.

[0119] Step a3: For each first trajectory point among the multiple first trajectory points, the controller generates a training sample and a category label corresponding to the training sample based on the trajectory data corresponding to the first trajectory point and the order information of multiple historical transportation orders.

[0120] It should be noted that for each first trajectory point, the controller uses the trajectory data corresponding to the first trajectory point (the position data of the first trajectory point, the vehicle speed data of the first trajectory point and the collection time point of the first trajectory point), as well as the order information of multiple historical transport orders, to create a training sample.

[0121] For each first trajectory point, the controller uses the loading point location data and the unloading point location data of the historical transport order to which the first trajectory point belongs, and assigns corresponding category labels to the training samples created based on the first trajectory point.

[0122] In this embodiment, based on the trajectory data corresponding to the first trajectory point and the order information of multiple historical transport orders, the generated training samples have high-quality feature data, and the category labels corresponding to the training samples accurately reflect the actual situation of the first trajectory point, ensuring the training effect of the initial loading and unloading point recognition model and improving the accuracy of the target loading and unloading point recognition model obtained by training.

[0123] In an embodiment of the present application, by acquiring and processing loading point location data, unloading point location data, transport vehicle model data, driver data and an initial trajectory point data set from multiple historical transport orders, it is possible to provide a comprehensive and high-quality training sample for the initial loading and unloading point recognition model, and generate corresponding category labels to ensure that the initial loading and unloading point recognition model can obtain accurate supervision information during the training process, thereby ensuring the training effect of the initial loading and unloading point recognition model and improving the accuracy of the target loading and unloading point recognition model obtained through training.

[0124] In the above embodiment, for each historical transport order among the multiple historical transport orders, the controller needs to determine the first trajectory point data set corresponding to the historical transport order based on the initial trajectory point data set corresponding to the historical transport order. Next, the specific process of the controller needing to determine the first trajectory point data set corresponding to the historical transport order based on the initial trajectory point data set corresponding to the historical transport order is described in detail.

[0125] In a possible embodiment, step a2, for each historical transport order among multiple historical transport orders, the controller determines the first trajectory point data set corresponding to the historical transport order based on the initial trajectory point data set corresponding to the historical transport order, which can be achieved through steps a21 to a23. Steps a21 to a23 are described in detail below.

[0126] Step a21, for each of the multiple historical transport orders, determine the average vehicle speed of the first target trajectory point based on the vehicle speed data of the fourth trajectory point, the collection time point of the fourth trajectory point, the vehicle speed data of the fifth trajectory point and the collection time point of the fifth trajectory point.

[0127] Among them, the first target trajectory point is any third trajectory point among multiple third trajectory points, the fourth trajectory point is a third trajectory point before the first target trajectory point among multiple third trajectory points, and the fifth trajectory point is a third trajectory point after the first target trajectory point among multiple third trajectory points.

[0128] It should be noted that for each historical transport order, first, any third trajectory point is selected from the initial trajectory point data set of the historical transport order as the first target trajectory point, and the fourth and fifth trajectory points are determined based on the first target trajectory point; second, the vehicle speed data and collection time point of the fourth trajectory point, as well as the vehicle speed data and collection time point of the fifth trajectory point are obtained; finally, according to the formula Calculate the average vehicle speed of the transport vehicle at the first target trajectory point

[0129] V 4 represents the instantaneous vehicle speed of the transport vehicle at the fourth trajectory point, V 5 represents the instantaneous vehicle speed of the transport vehicle at the fifth trajectory point, T 1 Indicates the time point when the first target trajectory point is collected, T 4 Indicates the time point when the fourth trajectory point is collected, T 5 Indicates the time point when the fifth trajectory point is collected.

[0130] It should be noted that, when the first target trajectory point is the first third trajectory point in the initial trajectory point data set, the instantaneous vehicle speed of the transport vehicle at the fourth trajectory point is zero, and the time point when the fourth trajectory point is collected is the time point when the first target trajectory point is collected; when the first target trajectory point is the last third trajectory point in the initial trajectory point data set, the instantaneous vehicle speed of the transport vehicle at the fifth trajectory point is zero, and the time point when the fifth trajectory point is collected is the time point when the first target trajectory point is collected.

[0131] In this embodiment, since the instantaneous vehicle speed may be affected by short-term, random factors, such as the driver's brief acceleration or deceleration, sudden changes in traffic conditions, etc., by calculating the average vehicle speed of the first target trajectory point, the actual driving behavior of the transport vehicle can be more accurately captured, which facilitates more accurate judgment of whether the first target trajectory point is the stop point of the target vehicle, and effectively improves the recognition accuracy of the first trajectory point.

[0132] Step a22: when the average vehicle speed at the first target trajectory point is zero, determine the first target trajectory point as the first trajectory point.

[0133] In this embodiment, whether the first target trajectory point is the first trajectory point is determined by judging whether the average vehicle speed of the first target trajectory point is zero. When the average vehicle speed of the first target trajectory point is zero, the first target trajectory point is determined to be the first trajectory point, so that the first trajectory points in the first trajectory point data set are all stop points of the transport vehicle, thereby ensuring the quality of the generated training samples.

[0134] Step a23: Based on the initial track point data set corresponding to the historical transport order and the first track point among the plurality of third track points, a first track point data set corresponding to the historical transport order is obtained.

[0135] It should be noted that after determining each first trajectory point among the multiple third trajectory points of the historical transport order, the trajectory data corresponding to the first trajectory point is filtered out from the initial trajectory point data set, and the trajectory data corresponding to all the first trajectory points are gathered into a set, which is the first trajectory point data set corresponding to the historical transport order.

[0136] In an embodiment of the present application, by calculating the average vehicle speed of each third trajectory point in the historical transport order, the first trajectory point (stop point) in the historical transport order is determined, and a first trajectory point data set is formed, which effectively reduces the error caused by instantaneous vehicle speed fluctuations and improves the training sample quality of the initial loading and unloading point recognition model, thereby enhancing the accuracy and reliability of the target loading and unloading point recognition model in practical applications.

[0137] In the above embodiment, for each of the plurality of first trajectory points, the controller needs to generate a training sample and a category label corresponding to the training sample based on the trajectory data corresponding to the first trajectory point and the order information of the plurality of historical transport orders. Next, the specific process of the controller generating the training sample and the category label corresponding to the training sample based on the trajectory data corresponding to the first trajectory point and the order information of the plurality of historical transport orders is described in detail.

[0138] It should be noted that for a first trajectory point, a training sample and a category label corresponding to the training sample can be generated. The method for generating multiple training samples and a category label corresponding to each training sample is the same. This embodiment only describes the method of generating a training sample and a category label corresponding to the training sample based on the trajectory data corresponding to a first trajectory point and the order information of multiple historical transportation orders.

[0139] In a possible embodiment, step a3, for each first trajectory point among multiple first trajectory points, the controller generates a training sample and a category label corresponding to the training sample based on the trajectory data corresponding to the first trajectory point and the order information of multiple historical transport orders. This can be achieved through steps a31 and a32. Steps a31 and a32 are described in detail below.

[0140] Step a31, the controller generates a training sample for each first trajectory point among multiple first trajectory points based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transport order, as well as the loading point location data, unloading point location data, transport vehicle model data and driver data of multiple historical transport orders.

[0141] The target historical transport order is the historical transport order to which the first track point belongs among the multiple historical transport orders.

[0142] It should be noted that the controller comprehensively analyzes the position data, vehicle speed data and collection time point of the first trajectory point, and organically combines the characteristics of the first trajectory point with the order information of multiple historical transport orders to generate training samples that can reflect the multi-dimensional information in the actual logistics transportation process. The training samples have rich feature description capabilities, provide more representative samples for the training of the initial loading and unloading point recognition model, and help improve the accuracy and robustness of the trained target loading and unloading point recognition model.

[0143] Step a32: The controller generates a category label corresponding to the training sample based on the position data and collection time point of the first trajectory point, and the loading point position data and unloading point position data of the target historical transportation order.

[0144] It should be noted that the controller compares the position of the first trajectory point with the loading point position and the unloading point position of the target historical transport order, and determines whether the first trajectory point is a loading point, an unloading point, or a stop point in combination with the collection time point of the first trajectory point, thereby determining the category label of the training sample to which the first trajectory point belongs. By labeling the category labels of the training samples, the initial loading and unloading point recognition model can better learn the characteristics of different categories of trajectory points during the training process, thereby improving the ability of the trained target loading and unloading point recognition model to distinguish between loading points, unloading points, and stop points in actual applications, and ultimately improving the accuracy and reliability of loading and unloading point recognition.

[0145] In an embodiment of the present application, by generating training samples for training the initial loading and unloading point recognition model and category labels corresponding to the training samples, it is ensured that the training samples used in the training of the initial loading and unloading point recognition model have high-quality features and accurate classification information, thereby providing a rich and real learning resource for the initial loading and unloading point recognition model. At the same time, it also effectively reduces the possibility of data noise and incorrect labeling, thereby enhancing the predictive performance and adaptability of the trained target loading and unloading point recognition model.

[0146] In the above embodiment, the controller needs to generate training samples based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transport order, as well as the loading point location data, unloading point location data, transport vehicle model data, and driver data of multiple historical transport orders. Next, the specific process of the controller generating training samples based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transport order, as well as the loading point location data, unloading point location data, transport vehicle model data, and driver data of multiple historical transport orders is described in detail.

[0147] In a possible embodiment, step a31, generating training samples based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transport order, as well as the loading point location data, unloading point location data, transport vehicle model data and driver data of multiple historical transport orders, can be implemented through steps a311 to a314, and steps a311 to a314 are described in detail below.

[0148] Step a311: The controller generates travel feature data of the first trajectory point based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transportation order.

[0149] The travel characteristic data of the first trajectory point includes vehicle speed related data, position related data, spatial data, steering angle data and detour ratio data of the first trajectory point. The travel characteristic data of the first trajectory point is used to describe the dynamic behavior of the transport vehicle before and after approaching the first trajectory point, and provides an identification basis for identifying different categories of first trajectory points.

[0150] It should be noted that the vehicle speed related data of the first trajectory point includes the instantaneous vehicle speed, average vehicle speed, instantaneous acceleration and average acceleration of multiple sixth trajectory points of the transport vehicle, and the instantaneous vehicle speed, average vehicle speed, instantaneous acceleration and average acceleration of multiple seventh trajectory points.

[0151] The multiple sixth trajectory points are multiple consecutive third trajectory points located before the first trajectory point in the initial trajectory point data set corresponding to the target historical transportation order, and the multiple seventh trajectory points are multiple consecutive third trajectory points located after the first trajectory point in the initial trajectory point data set corresponding to the target historical transportation order.

[0152] Based on the vehicle speed data of the sixth trajectory point in the initial trajectory point data set, the instantaneous vehicle speed of the transport vehicle at the sixth trajectory point can be obtained; the method for determining the average vehicle speed of the transport vehicle at the sixth trajectory point is the same as the method for determining the average vehicle speed of the first target trajectory point in the embodiment shown in the above step a21, and this embodiment will not be repeated here.

[0153] The instantaneous acceleration of the transport vehicle at any sixth trajectory point can be calculated according to the instantaneous acceleration formula: Calculated; a represents the instantaneous acceleration of the transport vehicle at the sixth trajectory point, V 2 represents the instantaneous vehicle speed of the transport vehicle at the sixth trajectory point, V 1 represents the instantaneous vehicle speed of the transport vehicle at the third trajectory point before the sixth trajectory point in the initial trajectory point data set, T 2 Indicates the time point when the sixth trajectory point is collected, T 1 Indicates the time point at which the third trajectory point located before the sixth trajectory point in the initial trajectory point data set is collected.

[0154] The average acceleration of the transport vehicle at any sixth trajectory point can be calculated according to the average acceleration formula: Calculated; represents the average acceleration of the transport vehicle at the sixth trajectory point, V 3 represents the instantaneous vehicle speed of the transport vehicle at the third trajectory point located one after the sixth trajectory point in the initial trajectory point data set, T 3 Indicates the time point at which the third trajectory point located after the sixth trajectory point in the initial trajectory point data set is collected.

[0155] The calculation method of the instantaneous vehicle speed, average vehicle speed, instantaneous acceleration and average acceleration of the seventh trajectory point is similar to that of the instantaneous vehicle speed, average vehicle speed, instantaneous acceleration and average acceleration of the sixth trajectory point, and will not be repeated in this embodiment.

[0156] It should be noted that the number of the plurality of sixth trajectory points and the plurality of seventh trajectory points related to the vehicle speed related data of the first trajectory point is equal, and the number of the plurality of sixth trajectory points and the plurality of seventh trajectory points related to the vehicle speed related data of the first trajectory point can be set by the model manager, and this embodiment does not specifically limit this. For example, the number of the sixth trajectory points and the number of the seventh trajectory points related to the vehicle speed related data of the first trajectory point are both 5.

[0157] In this embodiment, the vehicle speed related data of the first trajectory point can capture the speed change rate of the transport vehicle and help distinguish different types of stop points.

[0158] It should be noted that the location-related data of the first trajectory point includes the longitude and latitude of the first trajectory point, and the time point when the first trajectory point is collected.

[0159] It should be noted that the spatial data of the first track point refers to the spanning area formed by the transport vehicle between multiple sixth track points and the spanning area formed by the transport vehicle between multiple seventh track points.

[0160] Figure 2 A schematic diagram of the position relationship of multiple third trajectory points provided in an embodiment of the present application. Figure 2 As shown, the first track point is track point g1, the plurality of sixth track points are track point g61, track point g62, track point g63, track point g64 and track point g65 in the order of the acquisition time points, and the plurality of seventh track points are track point g71, track point g72, track point g73, track point g74 and track point g75 in the order of the acquisition time points. The spatial data of track point g1 refers to the spanning area of ​​the spanning region s6 formed by the track points g61, g62, g63, g64 and g65, and the spanning area of ​​the spanning region s7 formed by the track points g71, g72, g73, g74 and g75.

[0161] In this embodiment, the spatial data of the first trajectory point is used to analyze the stability or variability of the moving path of the transport vehicle around the first trajectory point, which can help distinguish different types of stop points.

[0162] It should be noted that the steering angle data of the first track point is used to describe the direction change of the transport vehicle when approaching or leaving the first track point. The steering angle data of the first track point includes the front and rear steering angles of the first track point and the front and rear center steering angles of the first track point.

[0163] The front and rear steering angles of the first trajectory point are the angles between the first connecting line and the second connecting line, the first connecting line is the connecting line between the center of the sixth trajectory point with the earliest collection time point among the multiple sixth trajectory points and the center of the first trajectory point, and the second connecting line is the connecting line between the center of the seventh trajectory point with the latest collection time point among the multiple seventh trajectory points and the center of the first trajectory point.

[0164] The front and rear center steering angle of the first trajectory point is the angle between the third line and the fourth line, the third line is the line connecting the center of the sixth trajectory point located in the middle among the multiple sixth trajectory points and the center of the first trajectory point, and the second line is the line connecting the center of the seventh trajectory point located in the middle among the multiple seventh trajectory points and the center of the first trajectory point.

[0165] like Figure 2 As shown, trajectory point g61 is the sixth trajectory point with the earliest collection time point among multiple sixth trajectory points, and the line connecting the center of trajectory point g61 and the center of trajectory point g1 is the first line L1; trajectory point g75 is the seventh trajectory point with the latest collection time point among multiple seventh trajectory points, and the line connecting the center of trajectory point g75 and the center of trajectory point g1 is the second line L2; the angle α between the first line L1 and the second line L2 is the front-rear steering angle of the first trajectory point.

[0166] Trajectory point g63 is the sixth trajectory point located in the middle among multiple sixth trajectory points, and the line connecting the center of trajectory point g63 and the center of trajectory point g1 is the third line L3; trajectory point g73 is the seventh trajectory point located in the middle among multiple seventh trajectory points, and the line connecting the center of trajectory point g73 and the center of trajectory point g1 is the fourth line L4; the angle β between the third line L3 and the fourth line L4 is the front and rear center steering angle of the first trajectory point.

[0167] In this embodiment, by calculating the front and rear steering angles and the front and rear center steering angles of the first trajectory point, the direction change of the transport vehicle when passing the first trajectory point can be described more accurately, providing rich direction change information for the feature data of the first trajectory point, which helps to better identify the behavior of the transport vehicle and enhance the recognition ability of the trained target loading and unloading point recognition model.

[0168] It should be noted that the detour ratio data of the first track point is used to measure whether the transport vehicle has made a detour when approaching and leaving the first track point. The detour ratio data of the first track point includes a front detour ratio and a rear detour ratio.

[0169] The front detour ratio of the first track point can be simplified by GeoHash coding. When it is detected that the GeoHash codes of multiple consecutive sixth track points are the same, the front detour ratio is directly defined as the number of consecutive sixth track points. For example, when it is detected that the GeoHash codes of three consecutive sixth track points are the same, the front detour ratio is determined to be 3.

[0170] The calculation method of the rear detour ratio and the front detour ratio of the first trajectory point is similar, and will not be described in detail in this embodiment.

[0171] The front detour ratio of the first trajectory point is used to detect whether the transport vehicle has taken a detour when approaching the first trajectory point. If the detour is relatively high, it may indicate that the transport vehicle is traveling in complex terrain or urban areas, and the first trajectory point may be a loading or unloading point; the rear detour ratio of the first trajectory point is used to detect whether the transport vehicle quickly returns to the main route after leaving the first trajectory point.

[0172] In this embodiment, the detour ratio data of the first trajectory point is used to evaluate whether there is a detour phenomenon when the transport vehicle passes through the first trajectory point. Through the front detour ratio data and the rear detour ratio data of the first trajectory point, the loading point, unloading point and stop point can be effectively distinguished, so as to enhance the recognition ability of the trained target loading and unloading point recognition model.

[0173] In an embodiment of the present application, by generating trip characteristic data including vehicle speed-related data, position-related data, spatial data, steering angle data and detour ratio data, the dynamic behavior of the transport vehicle at the first trajectory point can be fully described. The trip characteristic data provides multi-dimensional input for the initial loading and unloading point recognition model, which helps to accurately identify and classify the types of stop points, such as loading points, unloading points and rest points, thereby improving the recognition accuracy and prediction performance of the trained target loading and unloading point recognition model.

[0174] Step a312: The controller generates business feature data of the first trajectory point based on the location data of the first trajectory point, as well as the loading point location data, unloading point location data, transport vehicle model data and driver data of multiple historical transport orders.

[0175] The business characteristic data of the first trajectory point includes loading data, unloading data, transport vehicle type data and driver data of the first trajectory point.

[0176] It should be noted that the loading data of the first track point refers to the data of all loading operations related to the first track point within the target time period. The loading data of the first track point includes the total number of times loading is completed at the first track point within the target time period, and the total duration of loading at the first track point within the target time period.

[0177] In this embodiment, the total duration of loading at the first trajectory point within the target time period reflects the duration of the loading process, which can help the initial loading and unloading point recognition model identify the characteristics of the loading point; the total number of times loading is completed at the first trajectory point within the target time period can help the initial loading and unloading point recognition model determine whether the first trajectory point is a frequent loading point.

[0178] It should be noted that the unloading data of the first track point refers to the data of all unloading operations related to the first track point within the target time period. The unloading data of the first track point includes the total number of times unloading is completed at the first track point within the target time period, and the total duration of unloading at the first track point within the target time period.

[0179] In this embodiment, the total duration of unloading at the first trajectory point within the target time period reflects the duration of the unloading process, which can help the initial loading and unloading point recognition model identify the characteristics of the unloading point; the total number of times unloading is completed at the first trajectory point within the target time period can help the initial loading and unloading point recognition model determine whether the first trajectory point is a frequent unloading point.

[0180] It should be noted that the transport vehicle type data of the first trajectory point refers to the type and length of the transport vehicle used at the first trajectory point during the target time period. The transport vehicle type data provides information on vehicle type and length, which helps the initial loading and unloading point recognition model understand the impact of different vehicle types and lengths on loading and unloading operations.

[0181] It should be noted that the driver data of the first track point refers to the name, driving experience and ID number of the driver related to the first track point during the target time period. The driver data of the first track point is helpful to track the personnel related to the operation.

[0182] In this embodiment, the business characteristic data of the first trajectory point is formed by collecting and analyzing the loading data, unloading data, transport vehicle model data and driver data related to the first trajectory point. The business characteristic data of the first trajectory point can comprehensively describe the business activities and environmental characteristics of the first trajectory point, and can support the initial loading and unloading point recognition model to accurately classify the first trajectory point, thereby improving the recognition accuracy of the loading and unloading point by the trained target loading and unloading point recognition model.

[0183] Step a313: The controller fuses the travel feature data of the first trajectory point and the business feature data of the first trajectory point to obtain fused feature data of the first trajectory point.

[0184] It should be noted that the controller performs standardized splicing processing on the travel feature data of the first trajectory point and the business feature data of the first trajectory point to obtain the fused feature data of the first trajectory point.

[0185] In this embodiment, by fusing the travel feature data and business feature data of the first trajectory point, the situation of the first trajectory point can be fully described from multiple dimensions, which helps to more accurately capture the real features of the first trajectory point, avoid the deviation that may be caused by a single data source, and effectively improve the recognition accuracy of the initial loading and unloading point recognition model for different types of stop points (loading points, unloading points or stop points), so that the trained target loading and unloading point recognition model is more robust when dealing with complex situations. Step a314: The controller generates a training sample based on the fused feature data of the first trajectory point.

[0186] In an embodiment of the present application, high-quality training samples are generated by extracting and fusing the travel feature data and business feature data of the first trajectory point. The training samples include both the dynamic information of the transport vehicle at the first trajectory point and the data related to the business operations of the first trajectory point, thereby ensuring that the initial loading and unloading point recognition model can fully learn the characteristics of different types of stop points during the training process, so that the target loading and unloading point recognition model obtained through training has higher recognition accuracy and better generalization ability.

[0187] In the above embodiment, the controller needs to generate the category label corresponding to the training sample based on the position data and the collection time point of the first trajectory point, and the loading point position data and the unloading point position data of the target historical transport order. Next, the specific process of the controller generating the category label corresponding to the training sample based on the position data and the collection time point of the first trajectory point, and the loading point position data and the unloading point position data of the target historical transport order is described in detail.

[0188] In a possible embodiment, step a32, the controller generates category labels corresponding to the training samples based on the location data and collection time point of the first trajectory point, and the loading point location data and unloading point location data of the target historical transport order. This can be achieved through steps a321 to a325, and steps a321 to a325 are described in detail below.

[0189] Step a321: The controller determines the first distance based on the position data of the first trajectory point and the loading point position data of the target historical transport order.

[0190] It should be noted that the controller can calculate the straight-line distance between the first trajectory point and the loading point of the target historical transport order, i.e., the first distance, based on the longitude and latitude of the first trajectory point and the longitude and latitude of the loading point of the target historical transport order using the distance calculation formula between the two points.

[0191] Step a322: The controller determines the second distance based on the position data of the first trajectory point and the position data of the unloading point of the target historical transportation order.

[0192] It should be noted that the controller can calculate the straight-line distance between the first trajectory point and the unloading point of the target historical transport order, i.e., the second distance, based on the longitude and latitude of the first trajectory point and the longitude and latitude of the unloading point of the target historical transport order using the distance calculation formula between the two points.

[0193] Step a323: When the first distance is less than the first preset distance and the acquisition time point of the first trajectory point is within the first preset time period, the controller determines that the category label corresponding to the training sample is a loading point label.

[0194] When the category label corresponding to the training sample is the loading point label, the first trajectory point corresponding to the training sample is the loading point.

[0195] Among them, the first preset distance is the radius of the loading area corresponding to the loading point of the historical transport order.

[0196] It should be noted that the first preset distance can be set by the model manager himself, and this embodiment does not make any specific limitation to this.

[0197] For example, the first preset distance may be 1 kilometer.

[0198] The first preset time period is a time period between a first moment when the transport vehicle first enters the loading area and a second moment when the transport vehicle first leaves the loading area.

[0199] It should be noted that the controller can determine the collection time point (first moment) of the eighth trajectory point and the collection time point (second moment) of the ninth trajectory point by analyzing multiple third trajectory points corresponding to the target historical transport order and the loading point location data of the target historical transport order, and determine the time period between the collection time point (first moment) of the eighth trajectory point and the collection time point (second moment) of the ninth trajectory point as the first preset time period.

[0200] The eighth track point is the first third track point located within the loading area among the multiple third track points of the target historical transport order, and the ninth track point is the first third track point located outside the loading area after the eighth track point among the multiple third track points.

[0201] The first preset time period is determined based on real trajectory data rather than a preset static value, thereby improving the practical applicability and accuracy of the first preset time period and enhancing the recognition capability of the initial loading and unloading point recognition model for the loading point.

[0202] In this embodiment, by combining the geographical distance (first distance) between the first trajectory point and the loading point of the target historical transport order, as well as the collection time point of the first trajectory point, it is determined whether the first trajectory point is a loading point; the first trajectory point can be marked as a loading point only when the first distance is less than the first preset distance and the collection time point of the first trajectory point is within the first preset time period, thereby improving the accuracy of the category labels corresponding to the training samples.

[0203] Step a324: When the second distance is less than the second preset distance and the acquisition time point of the first trajectory point is within the second preset time period, the controller determines that the category label corresponding to the training sample is a unloading point label.

[0204] When the category label corresponding to the training sample is the unloading point label, the first trajectory point corresponding to the training sample is the loading point.

[0205] Among them, the second preset distance is the radius of the unloading area corresponding to the unloading point of the historical transport order.

[0206] It should be noted that the second preset distance can be set by the model manager himself, and this embodiment does not make any specific limitation to this.

[0207] For example, the second preset distance may be 1 kilometer.

[0208] The second preset time period is a time period between a third moment when the vehicle first enters the unloading area and a fourth moment when the vehicle first leaves the unloading area.

[0209] It should be noted that the controller can determine the collection time point (third moment) of the tenth trajectory point and the collection time point (fourth moment) of the eleventh trajectory point by analyzing multiple third trajectory points corresponding to the target historical transport order and the unloading point location data of the target historical transport order, and determine the time period between the collection time point (third moment) of the tenth trajectory point and the collection time point (fourth moment) of the eleventh trajectory point as the second preset time period.

[0210] The tenth track point is the first third track point located within the unloading area among the multiple third track points of the target historical transport order, and the eleventh track point is the first third track point located outside the unloading area after the tenth track point among the multiple third track points.

[0211] The second preset time period is determined based on real trajectory data rather than a preset static value, thereby improving the practical applicability and accuracy of the second preset time period and enhancing the recognition capability of the initial loading and unloading point recognition model for the unloading point.

[0212] In this embodiment, whether the first trajectory point is a unloading point is determined by combining the geographical distance (second distance) between the first trajectory point and the unloading point of the target historical transport order, as well as the collection time point of the first trajectory point; the first trajectory point can be marked as a unloading point only when the second distance is less than the second preset distance and the collection time point of the first trajectory point is within the second preset time period, thereby improving the accuracy of the category labels corresponding to the training samples.

[0213] Step a325: When the first preset condition is met, the controller determines that the category label corresponding to the training sample is a rest point label.

[0214] Among them, the first preset condition includes at least one of the following: the first distance is not less than the first preset distance, and the second distance is not less than the second preset distance; the first distance is less than the first preset distance, and the collection time point of the first trajectory point is outside the first preset time period; the second distance is less than the second preset distance, and the collection time point of the first trajectory point is outside the second preset time period.

[0215] It should be noted that, when the first trajectory point satisfies the first distance not less than the first preset distance and the second distance not less than the second preset distance, or the first distance is less than the first preset distance and the collection time point of the first trajectory point is outside the first preset time period, or the second distance is less than the second preset distance and the collection time point of the first trajectory point is outside the second preset time period, the first trajectory point is determined to be a rest point, that is, the category label corresponding to the training sample generated based on the first trajectory point is a rest point label.

[0216] In the embodiment of the present application, through precise geographic distance calculation and time period judgment, it is ensured that the category label of the training sample can accurately reflect the actual operation type of the first trajectory point, thereby improving the training quality and prediction accuracy of the initial loading and unloading point recognition model.

[0217] In the above embodiment, when the initial loading and unloading point identification model satisfies the model convergence condition, the target loading and unloading point identification model is obtained. Next, the model convergence condition is described in detail.

[0218] In a possible embodiment, the model convergence condition is that the loss function of the initial loading and unloading point identification model in consecutive target rounds is less than the target loss function.

[0219] The model convergence condition is used to determine when the training process of the initial loading and unloading point recognition model is stopped and to ensure that the initial loading and unloading point recognition model has achieved a satisfactory performance level during training.

[0220] The loss function is an indicator used to measure the gap between the prediction results of the initial loading and unloading point recognition model for the training samples and the category labels of the training samples.

[0221] The target loss function is a threshold pre-set by the model manager, which indicates the maximum loss value that the initial loading and unloading point identification model can accept during the training process.

[0222] The target number of consecutive rounds can be freely set by the model manager, and this embodiment does not specifically limit this. For example, the target number of consecutive rounds is 20 rounds.

[0223] In the embodiment of the present application, when the value of the loss function in each round of the training process of the initial loading and unloading point recognition model is less than the set target loss function value during the continuous target rounds, it means that the performance of the initial loading and unloading point recognition model in the continuous target rounds is stable and reaches a satisfactory level, and the training of the initial loading and unloading point recognition model can be stopped to obtain the target loading and unloading point recognition model. It can ensure that the initial loading and unloading point recognition model performs stably and as expected in the later stage of training, avoid overtraining or insufficient training, reduce the risk of overfitting, and ensure that the trained target loading and unloading point recognition model has good generalization ability.

[0224] In another possible embodiment, the model convergence condition is that the change in the loss function of the initial loading and unloading point identification model in the consecutive target number of rounds is less than the target loss function change.

[0225] The loss function change refers to the change in the loss function value between two adjacent rounds of training, that is, the loss function change is equal to the loss function value of the current round minus the loss function value of the previous round.

[0226] The target loss function change is a threshold value pre-set by the model manager, which indicates the minimum loss function change allowed for the initial loading and unloading point identification model during the training process.

[0227] In the embodiments of the present application,

[0228] When the loss function change of the initial loading and unloading point recognition model is less than the set target loss function change during the training of the continuous target number of rounds, it means that the improvement of the initial loading and unloading point recognition model is small and close to a stable state. At this time, the training of the initial loading and unloading point recognition model can be stopped to obtain the target loading and unloading point recognition model. It can ensure that the training of the initial loading and unloading point recognition model reaches a stable state and avoid unnecessary training rounds, which helps save computing resources and time and prevents the initial loading and unloading point recognition model from being over-adjusted during the training process, resulting in overfitting.

[0229] Figure 3 A schematic diagram of a method for identifying a loading and unloading point provided in an embodiment of the present application. Figure 3 As shown, the loading and unloading point identification method provided in the embodiment of the present application specifically includes S301 and S302, and S301 and S302 are described in detail below.

[0230] S301. Obtain a second trajectory point data set corresponding to a target transport order.

[0231] The second trajectory point data set includes trajectory data corresponding to a plurality of second trajectory points.

[0232] It should be noted that the target transport order may be an offline transport order or an online transport order that requires loading and unloading point identification.

[0233] S302. For each second trajectory point among the plurality of second trajectory points, input the fused feature data of the second trajectory point into a target loading and unloading point recognition model to obtain a prediction result output by the target loading and unloading point recognition model; wherein the prediction result is used to indicate a category of the second trajectory point.

[0234] The prediction result is used to indicate the category of the second trajectory point.

[0235] It should be noted that the prediction results output by the target loading and unloading point identification model are probability values ​​of three categories (resting point, loading point and unloading point). The probability value of each category represents the prediction confidence level of the target loading and unloading point identification model for each category.

[0236] The probability value of each category is a score between 0 and 1, and the sum of the probability values ​​of the three categories in the prediction results output by the target loading and unloading point recognition model is 1.

[0237] In one embodiment, the category corresponding to the maximum probability value in the prediction result output by the target loading and unloading point recognition model is determined as the category of the second trajectory point.

[0238] For example, the target loading and unloading point recognition model predicts the second trajectory point, and the output prediction results are shown in Table 1.

[0239] Table 1

[0240] category Loading point Unloading point Rest Point Probability 80％ 10％ 10％

[0241] In this case, in the prediction results output by the target loading and unloading point recognition model, the maximum probability value is 80%, and the corresponding category is the loading point. Therefore, the second trajectory point is determined to be the loading point.

[0242] In this embodiment, by selecting the category corresponding to the maximum probability value, the classification result of the second trajectory point can be clarified, ensuring that the category output by the model has the strongest prediction confidence, simplifying the classification decision process, and making the final category assignment clear and easy to explain.

[0243] In another embodiment, when the category corresponding to the largest probability value in the prediction results output by the target loading and unloading point identification model is a loading point or an unloading point, and when the probability value corresponding to the loading point or the unloading point in the prediction results is greater than a preset probability value, the second trajectory point is determined to be a loading point or an unloading point; when the probability value corresponding to the loading point or the unloading point in the prediction results is less than or equal to a preset probability value, the second trajectory point is determined to be a rest point.

[0244] The preset probability value may be set by the model management personnel, and this embodiment does not specifically limit this. For example, the preset probability value may be 70%.

[0245] For example, the target loading and unloading point recognition model predicts the second trajectory point, and the output prediction results are shown in Table 2.

[0246] Table 2

[0247] category Loading point Unloading point Rest Point Probability 65％ 25％ 10％

[0248] In this case, in the prediction results output by the target loading and unloading point identification model, the maximum probability value is 65%, and the corresponding category is loading point. Since the probability that the second trajectory point is a loading point in the prediction results output by the target loading and unloading point identification model is less than the preset probability value (70%), the second trajectory point is determined as a rest point.

[0249] For another example, the target loading and unloading point recognition model predicts the second trajectory point, and the output prediction result is shown in Table 3.

[0250] Table 3

[0251] category Loading point Unloading point Rest Point Probability 20％ 75％ 5％

[0252] In this case, in the prediction results output by the target loading and unloading point identification model, the maximum probability value is 75%, and the corresponding category is the unloading point. Since the probability that the second trajectory point is the unloading point in the prediction results output by the target loading and unloading point identification model is greater than the preset probability value (70%), the second trajectory point is determined to be the unloading point.

[0253] In this embodiment, by introducing a preset probability value, the classification result of the target loading and unloading point identification model not only takes into account the maximum probability value, but also ensures the confidence level of the prediction result, thereby reducing uncertainty.

[0254] Figure 4 A structural diagram of a training device for a loading and unloading point recognition model provided in an embodiment of the present application. Figure 4 As shown, the training device 400 for the loading and unloading point recognition model provided in this embodiment can exist independently and be used to implement the operations corresponding to the controller in the above-mentioned loading and unloading point recognition model training method embodiment.

[0255] The training device 400 for the loading and unloading point recognition model may include: a transceiver module 401 and a processing module 402. The processing module 402 is used for data processing, and the transceiver module 401 may implement corresponding communication functions. The transceiver module 401 may also be called a communication interface or a communication unit.

[0256] Optionally, the training device 400 for the loading and unloading point identification model may further include a storage unit, which may be used to store instructions and / or data, and the processing module 402 may read the instructions and / or data in the storage unit so that the training device 400 for the loading and unloading point identification model implements the steps implemented by the controller in the above-mentioned loading and unloading point identification model training method embodiment.

[0257] The transceiver module 401 is used to execute the receiving-related operations of the controller in the above-mentioned training method embodiment of the loading and unloading point recognition model, and the processing module 402 is used to execute the processing-related operations of the controller in the above-mentioned training method embodiment of the loading and unloading point recognition model.

[0258] Optionally, the transceiver module 401 may include a sending module and a receiving module. The sending module is used to perform the sending operation in the above-mentioned training method for the loading and unloading point recognition model. The receiving module is used to perform the receiving operation in the above-mentioned training method for the loading and unloading point recognition model.

[0259] It should be noted that the training device 400 for the loading and unloading point recognition model may include a sending module but not a receiving module. Alternatively, the training device 400 for the loading and unloading point recognition model may include a receiving module but not a sending module. Specifically, it depends on whether the above solution executed by the training device 400 for the loading and unloading point recognition model includes a sending action and a receiving action.

[0260] As an example, the training device 400 of the loading and unloading point recognition model is used to perform the above Figure 1 Actions performed by the controller in the illustrated embodiment.

[0261] The training device 400 for the loading and unloading point recognition model may include a processing module 402 .

[0262] Processing module 402 is used to obtain multiple training samples and category labels corresponding to the multiple training samples respectively; wherein each training sample includes fused feature data of the first trajectory point, the fused feature data of the first trajectory point includes travel feature data and business feature data of the first trajectory point, the travel feature data includes vehicle speed related data, position related data, spatial data, steering angle data and detour ratio data of the first trajectory point, and the business feature data includes loading data, unloading data, transport vehicle model data and driver data of the first trajectory point; the category label corresponding to each training sample is used to indicate the category of the first trajectory point, and the category label is a rest point label, a loading point label or an unloading point label.

[0263] The processing module 402 is further used to input a plurality of training samples into the initial loading and unloading point recognition model.

[0264] The processing module 402 is also used to train the initial loading and unloading point recognition model based on the category labels corresponding to multiple training samples, until the model convergence conditions are met, and obtain the target loading and unloading point recognition model; wherein the target loading and unloading point recognition model is used to predict the category of the second trajectory point, and the second trajectory point is any trajectory point corresponding to the target transport order.

[0265] It should be understood that the execution of the above-mentioned corresponding processes by each module has been described in detail in the above-mentioned training method embodiment of the loading and unloading point recognition model, and will not be repeated here for the sake of brevity.

[0266] The processing module 402 in the above embodiment can be implemented by at least one processor or processor-related circuit. The transceiver module 401 can be implemented by a transceiver or a transceiver-related circuit. The transceiver module 401 can also be called a communication unit or a communication interface. The storage unit can be implemented by at least one memory.

[0267] Figure 5 This is a schematic diagram of the structure of a loading and unloading point identification device provided in an embodiment of the present application. Figure 5 As shown, the loading and unloading point identification device 500 provided in this embodiment can exist independently and be used to implement the operations corresponding to the controller in the above loading and unloading point identification method embodiment.

[0268] The loading and unloading point identification device 500 may include: a transceiver module 501 and a processing module 502. The processing module 502 is used for data processing, and the transceiver module 501 may implement corresponding communication functions. The transceiver module 501 may also be called a communication interface or a communication unit.

[0269] Optionally, the loading and unloading point identification device 500 may further include a storage unit, which may be used to store instructions and / or data, and the processing module 502 may read the instructions and / or data in the storage unit so that the loading and unloading point identification device 500 implements the steps implemented by the controller in the above-mentioned loading and unloading point identification method embodiment.

[0270] The transceiver module 501 is used to perform the reception-related operations of the controller in the above-mentioned loading and unloading point identification method embodiment, and the processing module 502 is used to perform the processing-related operations of the controller in the above-mentioned loading and unloading point identification method embodiment.

[0271] Optionally, the transceiver module 501 may include a sending module and a receiving module. The sending module is used to perform the sending operation in the above-mentioned loading and unloading point identification method embodiment. The receiving module is used to perform the receiving operation in the above-mentioned loading and unloading point identification method embodiment.

[0272] It should be noted that the loading and unloading point identification device 500 may include a sending module but not a receiving module. Alternatively, the loading and unloading point identification device 500 may include a receiving module but not a sending module. Specifically, it depends on whether the above solution executed by the loading and unloading point identification device 500 includes a sending action and a receiving action.

[0273] As an example, the loading and unloading point identification device 500 is used to perform the above Figure 3 Actions performed by the controller in the illustrated embodiment.

[0274] The loading and unloading point identification device 500 may include a processing module 502 .

[0275] The processing module 502 is used to obtain a second trajectory point data set corresponding to the target transport order; wherein the second trajectory point data set includes trajectory data corresponding to a plurality of second trajectory points.

[0276] The processing module 502 is also used to input the fused feature data of the second trajectory point into the target loading and unloading point recognition model as claimed in any one of claims 1 to 7 for each second trajectory point among the multiple second trajectory points, so as to obtain a prediction result output by the target loading and unloading point recognition model; wherein the prediction result is used to indicate the category of the second trajectory point.

[0277] It should be understood that the execution of the above-mentioned corresponding processes by each module has been described in detail in the above-mentioned loading and unloading point identification method embodiment, and for the sake of brevity, it will not be repeated here.

[0278] The processing module 502 in the above embodiment can be implemented by at least one processor or processor-related circuit. The transceiver module 501 can be implemented by a transceiver or a transceiver-related circuit. The transceiver module 501 can also be called a communication unit or a communication interface. The storage unit can be implemented by at least one memory.

[0279] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 6 As shown, the electronic device 600 provided in this embodiment includes: a memory 601 and a processor 602 .

[0280] The memory 601 may be an independent physical unit, and may be connected to the processor 602 via a bus 603. The memory 601 and the processor 602 may also be integrated together and implemented by hardware. The memory 601 is used to store program instructions, and the processor 602 calls the program instructions to execute the operations executed by the controller in any of the above method embodiments.

[0281] Optionally, when part or all of the methods of the above embodiments are implemented by software, the electronic device 600 may also include only a processor 602. The memory 601 for storing programs is located outside the electronic device 600, and the processor 602 is connected to the memory through circuits / wires for reading and executing the programs stored in the memory. The processor 602 may be a central processing unit (CPU), a network processor (NP) or a combination of a CPU and a NP. The processor 602 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD) or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL) or any combination thereof.

[0282] The memory 601 may include a volatile memory, such as a random-access memory (RAM); the memory may also include a non-volatile memory, such as a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD); the memory may also include a combination of the above types of memory.

[0283] Exemplarily, the present application provides a chip, including: an interface circuit and a logic circuit, the interface circuit is used to receive signals from other chips outside the chip and transmit them to the logic circuit, or send signals from the logic circuit to other chips outside the chip, and the logic circuit is used to execute the operations performed by the controller in the above method embodiment.

[0284] Illustratively, the present application provides a computer-readable storage medium on which computer program instructions are stored. The computer program instructions are executed by a processor of an electronic device so that the electronic device executes the operations performed by the controller in the above method embodiment.

[0285] Illustratively, the present application provides a computer program product. When the computer program product is executed on an electronic device, the electronic device executes the operations executed by the controller in the above method embodiment.

[0286] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A training method for a loading and unloading point recognition model, characterized in that: The training method comprises: Acquire multiple training samples and category labels corresponding to the multiple training samples respectively; wherein each of the training samples includes fused feature data of a first trajectory point, the fused feature data of the first trajectory point includes travel feature data and business feature data of the first trajectory point, the travel feature data includes vehicle speed related data, position related data, spatial data, steering angle data and detour ratio data of the first trajectory point, and the business feature data includes loading data, unloading data, transport vehicle model data and driver data of the first trajectory point; the category label corresponding to each of the training samples is used to indicate the category of the first trajectory point, and the category label is a rest point label, a loading point label or an unloading point label; Inputting the plurality of training samples into an initial loading and unloading point recognition model; Based on the category labels corresponding to the multiple training samples, the initial loading and unloading point recognition model is trained until the model convergence condition is met, so as to obtain a target loading and unloading point recognition model; wherein the target loading and unloading point recognition model is used to predict the category of the second trajectory point, and the second trajectory point is any trajectory point corresponding to the target transport order.

2. The method according to claim 1, characterized in that Before obtaining a plurality of training samples and the category labels respectively corresponding to the plurality of training samples, the method further includes: Acquire order data of multiple historical transport orders; wherein the order data of each of the historical transport orders includes loading point location data, unloading point location data, transport vehicle model data and driver data of the historical transport order, and an initial trajectory point data set corresponding to the historical transport order; the initial trajectory point data set includes trajectory data corresponding to multiple third trajectory points respectively, and the trajectory data corresponding to each of the third trajectory points includes the location data of the third trajectory point, vehicle speed data and collection time point; For each of the plurality of historical transport orders, based on the initial track point data set corresponding to the historical transport order, determine a first track point data set corresponding to the historical transport order; wherein the first track point data set includes track data corresponding to a plurality of first track points respectively; For each first trajectory point of the plurality of first trajectory points, the training sample and the category label corresponding to the training sample are generated based on the trajectory data corresponding to the first trajectory point and the order information of the plurality of historical transportation orders.

3. The method according to claim 2, characterized in that The step of determining, for each of the plurality of historical transport orders, a first track point data set corresponding to the historical transport order based on the initial track point data set corresponding to the historical transport order, comprises: For each of the plurality of historical transport orders, determining an average vehicle speed of a first target trajectory point according to vehicle speed data of a fourth trajectory point, a collection time point of the fourth trajectory point, vehicle speed data of a fifth trajectory point, and a collection time point of the fifth trajectory point; wherein the first target trajectory point is any third trajectory point among the plurality of third trajectory points, the fourth trajectory point is a third trajectory point among the plurality of third trajectory points that is located before the first target trajectory point, and the fifth trajectory point is a third trajectory point among the plurality of third trajectory points that is located after the first target trajectory point; In a case where an average vehicle speed of the first target trajectory point is zero, determining the first target trajectory point as the first trajectory point; Based on the initial track point data set corresponding to the historical transport order and the first track point among the plurality of third track points, a first track point data set corresponding to the historical transport order is obtained.

4. The method according to claim 2, characterized in that: The step of generating, for each of the plurality of first trajectory points, the training sample and the category label corresponding to the training sample based on the trajectory data corresponding to the first trajectory point and the order information of the plurality of historical transport orders comprises: For each of the plurality of first trajectory points, the training sample is generated based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transport order, as well as the loading point location data, unloading point location data, transport vehicle model data and driver data of the plurality of historical transport orders; wherein the target historical transport order is the historical transport order to which the first trajectory point belongs among the plurality of historical transport orders; Based on the position data and collection time point of the first trajectory point, and the loading point position data and unloading point position data of the target historical transportation order, a category label corresponding to the training sample is generated.

5. The method according to claim 4, characterized in that The generating of the training sample based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transport order, as well as the loading point location data, unloading point location data, transport vehicle model data and driver data of the plurality of historical transport orders comprises: Generate travel feature data of the first trajectory point based on the trajectory data corresponding to the first trajectory point and the initial trajectory point data set corresponding to the target historical transportation order; Generate business feature data of the first trajectory point based on the location data of the first trajectory point, and the loading point location data, unloading point location data, transportation vehicle model data, and driver data of the plurality of historical transportation orders; fusing the travel feature data of the first trajectory point and the business feature data of the first trajectory point to obtain fused feature data of the first trajectory point; The training sample is generated based on the fused feature data of the first trajectory point.

6. The method according to claim 4, characterized in that The step of generating a category label corresponding to the training sample based on the location data and the collection time point of the first trajectory point and the loading point location data and the unloading point location data of the target historical transport order includes: determining a first distance based on the location data of the first trajectory point and the location data of the loading point of the target historical transport order; determining a second distance based on the location data of the first trajectory point and the location data of the unloading point of the target historical transport order; When the first distance is less than a first preset distance and the collection time point of the first trajectory point is within a first preset time period, the category label corresponding to the training sample is determined to be a loading point label; wherein the first preset distance is the radius of the loading area corresponding to the loading point of the historical transport order, and the first preset time period is the time period between a first moment when the vehicle first enters the loading area and a second moment when the vehicle first leaves the loading area; When the second distance is less than the second preset distance and the collection time point of the first trajectory point is within the second preset time period, the category label corresponding to the training sample is determined to be the unloading point label; wherein the second preset distance is the radius of the unloading area corresponding to the unloading point of the historical transport order, and the second preset time period is the time period between the third moment when the transport vehicle first enters the unloading area and the fourth moment when the transport vehicle first leaves the unloading area; When the first preset condition is met, determining that the category label corresponding to the training sample is a rest point label; The first preset condition includes at least one of the following: The first distance is not less than the first preset distance, and the second distance is not less than the second preset distance; The first distance is smaller than the first preset distance, and the collection time point of the first trajectory point is outside the first preset time period; The second distance is smaller than the second preset distance, and the collection time point of the first trajectory point is outside the second preset time period.

7. The method according to claim 1, characterized in that The model convergence condition is that the loss function of the initial loading and unloading point identification model in the continuous target number of rounds is less than the target loss function; or, The change in the loss function of the initial loading and unloading point identification model in the continuous target number of rounds is less than the target loss function change.

8. A method for identifying a loading and unloading point, characterized in that: The method comprises: Acquire a second trajectory point data set corresponding to the target transport order; wherein the second trajectory point data set includes trajectory data corresponding to a plurality of second trajectory points respectively; For each of the multiple second trajectory points, the fused feature data of the second trajectory point is input into the target loading and unloading point recognition model as described in any one of claims 1 to 7 to obtain a prediction result output by the target loading and unloading point recognition model; wherein the prediction result is used to indicate the category of the second trajectory point.

9. An electronic device, characterized in that: include: memory and at least one processor; The memory stores computer-executable instructions; The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the method according to any one of claims 1 to 7, or the method according to claim 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed, the method according to any one of claims 1 to 7, or the method according to claim 8, is implemented.

Images