A logistics cloud platform information transmission monitoring method and device based on big data

Abstract

The application relates to the technical field of logistics supervision, and discloses a logistics cloud platform information transmission supervision method and device based on big data, which comprises the following steps: receiving sensor data flow of a target carrying vehicle; dividing a plurality of dynamic driving scene segments according to the sensor data; determining a target dynamic driving scene segment at a current time, and constructing a multi-dimensional baseline feature statistical model; inputting a data point at the current time into the multi-dimensional baseline feature statistical model, and calculating an abnormal score of the data point; when the abnormal score is greater than a preset risk alarm score, generating a risk event record; generating an early warning instruction based on the risk event record, and sending the early warning instruction to a logistics monitoring center and a vehicle-mounted terminal of the target carrying vehicle. According to the embodiment of the application, the division between normal and abnormal can be automatically adjusted according to specific driving scenes, false positives and false negatives caused by a fixed threshold can be significantly eliminated, accurate monitoring of dynamic driving behavior can be truly realized, and the original monitoring blind area is eliminated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of logistics supervision, and particularly relates to a logistics cloud platform information transmission supervision method and device based on big data. BACKGROUND

[0002] With the rapid development of e-commerce and intelligent manufacturing, the cross-regional transportation volume of bulk goods such as vehicle consignments continues to rise, and the coverage of the logistics network continues to expand. How to efficiently collect, transmit and analyze the real-time position, environmental state and safety alarm information of consigned vehicles has become a core requirement for ensuring vehicle safety and improving transportation efficiency.

[0003] In the prior art, the risk identification and early warning function during vehicle driving is usually realized through an alarm system based on a preset fixed threshold. For example, an alarm is triggered when the instantaneous acceleration or speed of the vehicle exceeds a fixed limit. However, the method based on a fixed threshold in the prior art has the risk of false positives or false negatives for normal driving behavior and abnormal driving behavior under different road environments in actual application. SUMMARY

[0004] Therefore, the present application provides a logistics cloud platform information transmission supervision method and device based on big data to solve the above problems.

[0005] In a first aspect, a logistics cloud platform information transmission supervision method based on big data is provided, applied to a server, and the method comprises the following steps.

[0006] Continuously receiving sensor data stream uploaded by the vehicle terminal of the target consigned vehicle, the sensor data stream containing a plurality of data points; processing the sensor data stream, identifying and dividing a plurality of continuous dynamic driving scene segments;

[0007] Determining the target dynamic driving scene segment at the current time, selecting a preset number of data points starting from the starting time of the target dynamic driving scene segment as modeling samples, and constructing a multi-dimensional baseline feature statistical model for the target dynamic driving scene segment based on the modeling samples;

[0008] Inputting the first data point at the current time into the multi-dimensional baseline feature statistical model to calculate the abnormal score of the first data point;

[0009] When the abnormal score is greater than a preset risk alarm score, generating a risk event record based on the first data point and the abnormal score, generating an early warning instruction based on the risk event record, and sending the early warning instruction to the logistics monitoring center and the vehicle terminal of the target consigned vehicle.

[0010] The above technical scheme automatically divides continuous segments (such as rapid acceleration, frequent lane changing, rough road conditions, etc.) that truly reflect the driving behavior state by performing clustering analysis on vehicle sensor data, and then extracts modeling samples from each segment to construct a baseline model for adaptive modeling of the "normal fluctuation range" under different driving scenarios. Compared with a fixed threshold, this modeling method based on scenario learning can dynamically adapt to changes in the vehicle operating state. The introduced abnormal score mechanism can compare and evaluate each new data point with the "scenario model", thereby accurately identifying potential risky behaviors (such as abnormal acceleration, skewed jolting, unstable posture, etc.) that deviate from the current state, and prompting the logistics monitoring center and the vehicle terminal to respond in real time, forming a closed loop. The division between normal and abnormal is automatically adjusted according to the specific driving scenario, significantly reducing false positives and false negatives caused by fixed thresholds, and truly achieving accurate monitoring of dynamic driving behavior, eliminating the original monitoring blind spot.

[0011] Optionally, continuously receiving the sensor data stream uploaded by the vehicle terminal of the target carrying vehicle comprises:

[0012] Receiving the original data packet uploaded by the target carrying vehicle, the original data packet containing three-axis acceleration, instantaneous speed and position information; analyzing the original data packet, and aligning the three-axis acceleration, instantaneous speed and position information in time sequence according to the time stamp to obtain multi-dimensional time sequence data;

[0013] Filtering the multi-dimensional time sequence to obtain the sensor data stream.

[0014] The above technical scheme analyzes, time-aligns and filters the original data packet uploaded by the vehicle terminal, so that the server obtains structured and reliable multi-dimensional time sequence data, thereby ensuring the consistency and accuracy of the basic data for subsequent clustering modeling and anomaly analysis. The time sequence distortion and feature imbalance problems caused by time stamp errors and data noise are eliminated, providing high-quality, clean and continuous input data for subsequent dynamic scenario identification, and improving the model robustness and analysis accuracy.

[0015] Optionally, processing the sensor data stream to identify and divide a plurality of continuous dynamic driving scenario segments comprises:

[0016] Dividing the sensor data stream into a plurality of clustering analysis windows according to a preset time length;

[0017] Calculating a window feature vector of each clustering analysis window, the window feature vector including a window acceleration standard deviation, a window speed change rate and a window tilt angle fluctuation rate;

[0018] Processing the window feature vector by a clustering analysis algorithm to obtain a cluster label, and assigning the cluster label to the clustering analysis window corresponding to the window feature vector;

[0019] adjacent clustering analysis windows with the same cluster label and adjacent time intervals less than a preset threshold are merged to generate continuous dynamic driving scene segments.

[0020] The above technical solution can realize adaptive division of continuous behavior patterns in a real driving state of a vehicle by dividing a sensor data stream into clustering analysis windows and performing density clustering based on feature indicators (acceleration standard deviation, speed change rate, and tilt angle fluctuation rate) in the windows. The clustering method enables the system to accurately identify subdivided scenes such as smooth cruising, rapid acceleration, and continuous vibration, thereby providing semantic support for constructing baseline models corresponding to the scenes. The context accuracy of risk assessment is significantly improved, and false positives or false negatives caused by mixed sample modeling are avoided.

[0021] Optionally, a target dynamic driving scene segment at the current time is determined, a preset number of data points from the start time of the target dynamic driving scene segment are selected as modeling samples, and a multi-dimensional baseline feature statistical model is constructed for the target dynamic driving scene segment based on the modeling samples, including:

[0022] A multi-dimensional feature corresponding to a second data point in the modeling samples is calculated to form a feature vector of the second data point, the second data point being any data point in the modeling samples, and the feature vector including acceleration standard deviation, speed change rate, and tilt angle fluctuation rate;

[0023] All feature vectors in the modeling samples are combined into a feature vector set;

[0024] Statistical distribution parameters of the distribution of the feature vector set in the multi-dimensional feature space are calculated;

[0025] The multi-dimensional baseline feature statistical model is constructed based on the statistical distribution parameters.

[0026] The above technical solution realizes data-driven modeling of the boundaries of normal behavior characteristics in different driving states by extracting representative modeling samples for each dynamic driving scene segment and constructing a statistical model (represented in the form of distribution parameters) containing multi-dimensional features. The model has dynamic adaptability and can form a standard template with exclusive discrimination ability for different segments. Compared with using uniform rules or fixed thresholds, scene feature modeling can significantly improve the accuracy of abnormal identification and scene perception ability, accurately distinguish normal jitter from real abnormalities based on scene features, improve the model discrimination ability, and reduce the false positive probability.

[0027] Optionally, the first data point at the current time is input into the multi-dimensional baseline feature statistical model, and an abnormal score of the first data point is calculated, including:

[0028] The multi-dimensional features of the first data point at the current time are extracted to generate a target feature vector;

[0029] inputting the target feature vector into the multi-dimensional baseline feature statistical model, calculating a probability that the target feature vector conforms to a normal driving behavior baseline defined by the multi-dimensional baseline feature statistical model by using the statistical distribution parameters;

[0030] generating an anomaly value for the first data point according to the probability, wherein the lower the probability, the higher the anomaly value;

[0031] normalizing the anomaly value to obtain an anomaly score.

[0032] The above technical solution realizes point-scenario-model integrated evaluation by extracting a multi-dimensional feature vector of a current target data point, calculating a conformity probability in combination with baseline statistical distribution parameters corresponding to the scenario, and normalizing the anomaly score. This mechanism can quantify the anomaly degree of each point and is highly associated with the current driving scenario. Accurate risk indicators can be given at the moment of an event, making the early warning more timely and more interpretable.

[0033] Optionally, when the anomaly score is greater than a preset risk warning score value, a risk event record is generated based on the first data point and the anomaly score, including:

[0034] When the anomaly score exceeds the preset risk warning score value, the first data point and the corresponding target feature vector are locked;

[0035] extracting target location information from the first data point;

[0036] mapping the anomaly score to a preset risk level based on a numerical interval in which the anomaly score is located;

[0037] analyzing numerical values of each dimension of the target feature vector, calculating a contribution weight of each dimension to the anomaly score, and determining a dimension with the highest contribution weight as a main anomaly dimension;

[0038] determining a risk type by matching a preset risk type mapping table according to the main anomaly dimension and a numerical deviation degree of the main anomaly dimension, wherein the numerical deviation degree is a deviation amplitude of a numerical value of the main anomaly dimension from the normal driving behavior baseline;

[0039] combining the target location information, the risk level, and the risk type to generate the risk event record.

[0040] The above technical solution records the anomaly score and combines location information, a risk level mapped by the score, and a contribution weight of each feature dimension to locate the root cause of the anomaly through a risk type mapping table. This structured generation method makes the risk event traceable and classifiable, supports subsequent precise intervention and statistical analysis, and significantly improves the accident traceability efficiency and decision support capability.

[0041] Optionally, the risk event record is used to generate a warning instruction, and the warning instruction is sent to a logistics monitoring center and a vehicle terminal of the target delivery vehicle, including:

[0042] Target position information, a risk level, and a risk type are parsed from the risk event record;

[0043] According to the risk type and the target position information, a warning content is generated;

[0044] According to the risk level, one or more warning forms are selected from a preset warning strategy library, and the warning forms include a sound alarm, a visual highlight prompt, or a text message;

[0045] The warning content and the warning form are encapsulated into a warning instruction;

[0046] The warning instruction is synchronously sent to the logistics monitoring center and the vehicle terminal of the target delivery vehicle.

[0047] The above technical solution encapsulates the risk event information into a standardized warning instruction and sends it to the logistics monitoring center and the vehicle terminal, respectively, to realize a front-end and back-end collaborative response mechanism. The intelligent analysis result can be converted into an execution signal in the first time to form a closed-loop path of "discovery-reporting-response", improve the emergency response efficiency, reduce potential safety accidents, and enhance the transparency and reliability of the logistics system operation.

[0048] In a second aspect of the present application, a logistics cloud platform information transmission monitoring system based on big data is provided, which includes a data acquisition module, a segment division module, a model construction module, a score calculation module, a risk record module, and a warning generation module, wherein:

[0049] The data acquisition module is configured to continuously receive sensor data streams uploaded by a vehicle terminal of a target delivery vehicle, and the sensor data streams include multiple data points;

[0050] The segment division module is configured to process the sensor data streams, identify and divide continuous multiple dynamic driving scene segments;

[0051] The model construction module is configured to determine a target dynamic driving scene segment at the current time, select a preset number of data points starting from the beginning of the target dynamic driving scene segment as a modeling sample, and construct a multi-dimensional baseline feature statistical model for the target dynamic driving scene segment based on the modeling sample;

[0052] The score calculation module is configured to input a first data point at the current time into the multi-dimensional baseline feature statistical model to calculate an abnormal score of the first data point;

[0053] a risk record module configured to generate a risk event record based on the first data point and the anomaly score when the anomaly score is greater than a preset risk alarm score;

[0054] a warning generation module configured to generate a warning instruction based on the risk event record and send the warning instruction to a logistics monitoring center and a vehicle terminal of the target delivery vehicle.

[0055] In a third aspect of the present application, an electronic device is provided, comprising a processor, a memory, a user interface and a network interface, the memory is configured to store instructions, the user interface and the network interface are configured to communicate with other devices, and the processor is configured to execute the instructions stored in the memory to enable the electronic device to perform the method according to any one of the preceding aspects.

[0056] In a fourth aspect of the present application, a computer readable storage medium is provided, which stores instructions, when the instructions are executed, the method according to any one of the preceding aspects is performed.

[0057] In summary, the one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0058] By clustering analysis on vehicle sensor data, continuous segments (such as sudden acceleration, frequent lane changing, rough road, etc.) that truly reflect the driving behavior state are automatically divided, and then modeling samples are extracted from each segment to construct a baseline model for adaptive modeling of the "normal fluctuation range" under different driving scenarios. Compared with the fixed threshold, this modeling method based on scenario learning can dynamically adapt to the changes of the vehicle operating state. The introduced anomaly score mechanism can compare and evaluate each new data point with the "scenario model", so as to accurately identify potential risk behaviors (such as abnormal acceleration, abnormal oscillation, unstable posture, etc.) that are different from the current state, and push the logistics monitoring center and the vehicle terminal to respond in real time, forming a closed loop. The division between normal and abnormal is automatically adjusted according to the specific driving scenario, significantly eliminating false positives and false negatives caused by the fixed threshold, and truly realizing accurate monitoring of dynamic driving behavior, and eliminating the original monitoring blind area. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 is an exemplary system architecture diagram of applying the information transmission monitoring method or the information transmission monitoring system of the logistics cloud platform based on big data of the present application;

[0060] Figure 2 is a flowchart of the information transmission monitoring method of the logistics cloud platform based on big data in the embodiments of the present application;

[0061] Figure 3is a module schematic diagram of a logistics cloud platform information transmission supervision system based on big data in the embodiment of the present application.

[0062] Figure 4 is a structural schematic diagram of an electronic device disclosed in the embodiment of the present application.

[0063] Legend: 100, system architecture; 101, first terminal device; 102, second terminal device; 103, third terminal device; 104, network; 105, server; 301, data acquisition module; 302, segment division module; 303, model construction module; 304, score calculation module; 305, risk record module; 306, early warning generation module; 401, processor; 402, communication bus; 403, user interface; 404, network interface; 405, memory. DETAILED DESCRIPTION

[0064] In order for those skilled in the art to better understand the technical solutions in the specification, the technical solutions in the embodiments of the specification will be clearly and completely described below in conjunction with the drawings in the embodiments of the specification. Obviously, the described embodiments are only some of the embodiments of the present application, not all.

[0065] In the description of the embodiments of the present application, the words such as "for example" or "for instance" are used to represent an example, illustration or description. Any embodiment or design scheme described as "for example" or "for instance" in the embodiments of the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the words such as "for example" or "for instance" are intended to present the relevant concept in a specific manner.

[0066] In the description of the embodiments of the present application, the term "a plurality of" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. In addition, the terms "first" and "second" are used for description purposes only, and should not be interpreted as indicating or implying relative importance or implicitly indicating the indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more of the features. The terms "include", "contain", "have" and their variants mean "include but are not limited to", unless otherwise specifically emphasized.

[0067] The embodiments disclose a logistics cloud platform information transmission supervision method based on big data or a logistics cloud platform information transmission supervision system based on big data, Figure 1An exemplary system architecture diagram of an embodiment of a logistics cloud platform information transmission monitoring method or system based on big data to which the present application can be applied is shown.

[0068] As shown in Figure 1 The system architecture 100 can include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104 and a server 105. The network 104 is a medium for providing a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 can include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0069] The user can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the terminal devices 101, 102, 103, such as model training applications, video recognition applications, web browser applications, social platform software, etc.

[0070] The terminal devices 101, 102, 103 can be hardware or software. When the terminal devices 101, 102, 103 are hardware, they can be various electronic devices with a display screen, including but not limited to smartphones, tablet computers, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers and desktop computers, etc. When the terminal devices 101, 102, 103 are software, they can be installed in the above-mentioned electronic devices. They can be implemented as multiple software or software modules (such as multiple software or software modules for providing distributed services) or as a single software or software module. No specific limitation is made herein.

[0071] When the terminal 101, 102, 103 is hardware, a video capture device can also be installed thereon. The video capture device can be various devices capable of capturing video, such as a camera, a sensor, etc. The user can use the video capture device on the terminal 101, 102, 103 to capture video.

[0072] The server 105 can be a server providing various services, such as a background server for processing data displayed on the terminal devices 101, 102, 103. The background server can analyze and process received data, and can feed back the processing result (such as a recognition result) to the terminal device.

[0073] It should be noted that the server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When the server is software, it can be implemented as multiple software or software modules (such as multiple software or software modules for providing distributed services), or as a single software or software module. No specific limitation is made herein.

[0074] It should be understood that Figure 1 The number of terminal devices, networks and servers in the system architecture is merely illustrative. Any number of terminal devices, networks and servers can be provided according to implementation needs. In particular, in the case where target data does not need to be acquired remotely, the above system architecture can not include a network, but only include a terminal device or a server.

[0075] Figure 2 is a flowchart of a logistics cloud platform information transmission monitoring method based on big data in an embodiment of the present application, as shown in Figure 2 The embodiment includes the following steps:

[0076] In step S201, sensor data stream uploaded by the vehicle terminal of the target delivery vehicle is continuously received, and the sensor data stream includes multiple data points.

[0077] Exemplarily, the MCU (Microcontroller Unit) of the vehicle terminal collects sensor data at different frequencies, 100Hz three-axis acceleration, 1Hz instantaneous speed and position information are aggregated and packaged into a JSON (JavaScript Object Notation) format raw data packet every 1 second, and uploaded to the cloud platform through the MQTT (Message Queuing Telemetry Transport) protocol; the cloud platform first verifies and parses the data packet, associates the speed and position information with the reference timestamp, reduces the dimension of the high-frequency acceleration data to aggregate and extract key statistical values, and then uses a Kalman filter to correct the position and speed, and a moving average filter to smooth the acceleration peak value, finally generates a time-aligned and dimension-unified sensor data stream and stores it.

[0078] In one possible implementation, the sensor data stream uploaded by the on-board terminal of the target delivery vehicle is continuously received, specifically including: receiving the original data packet uploaded by the target delivery vehicle, the original data packet containing three-axis acceleration, instantaneous speed and position information; parsing the original data packet, and time series aligning the three-axis acceleration, instantaneous speed and position information according to the time stamp to obtain multi-dimensional time series data; filtering the multi-dimensional time series to obtain the sensor data stream.

[0079] Specifically, the MCU in the on-board terminal collects sensor data at different frequencies: for example, three-axis acceleration, with a collection frequency of 100 Hz (100 data points per second) to capture high-frequency vibrations and impacts. The X-axis represents the acceleration / deceleration of the vehicle in the forward / reverse direction, the Y-axis represents the shaking / tilting in the left / right direction, and the Z-axis represents the jolting / impact in the vertical direction. Instantaneous speed: collection frequency of 1 Hz (1 data point per second), read through the CAN (Controller Area Network) bus. Position information: collection frequency of 1 Hz, obtain latitude, longitude, altitude and UTC timestamp. In order to optimize transmission efficiency, the on-board terminal does not send each data point separately, but packages all the data collected in 1 second into an original data packet (format: JSON), and then uploads it to the cloud platform through a pre-set communication protocol (for example, MQTT protocol).

[0080] Further, the data processing microservice of the cloud platform receives the above-mentioned raw data packet and immediately performs parsing and alignment. For example, the application first verifies the integrity and format of the data packet (such as whether the deviceid exists, whether the JSON format is correct), and then parses the JSON string into a structured object in memory, facilitating subsequent access to each field. The timestamp in the data packet is used as the reference timestamp of this second, and since the collection frequency of instantaneous speed and position information is consistent with the data packet sending frequency (both 1 Hz), it can be directly associated with the reference timestamp. The three-axis acceleration is a high-frequency data stream (100 Hz), and directly storing all 100 data points will cause data redundancy and difficulty in analysis, so it needs to be reduced for aggregation processing to extract key statistical values that can reflect the vibration characteristics in this second. For example, traverse the three-axis acceleration to find the maximum value of the Z-axis acc_z_max (such as 12.5 m / s), which may indicate that the vehicle has passed a speed bump or pothole. Calculate the average values of X-axis, Y-axis and Z-axis readings acc_x_avg, acc_y_avg, acc_z_avg to get the average vibration of the vehicle in each direction. Calculate the standard deviation of the Z-axis readings acc_z_stddev to get the severity of the vehicle's vertical jolt. Through parsing and alignment, a structured and dimensionally uniform multi-dimensional time series is generated and stored in a time series database (such as InfluxDB or TinmescaleDB).

[0081] Further, the multi-dimensional time series is filtered. For example, for position information and instantaneous speed, a Kalman filter can be used to predict the current state based on the previous position / speed state, and then combine the current position information and instantaneous speed measurement to obtain an optimal estimate. For the aggregated three-axis acceleration feature data, a moving average filter is used, for example, a moving average with a window size of 5 is applied to the acc_z_max time series, such as T (current) second acc_z_max = 12.5, T-1 second acc_z_max = 10.3, T-2 second acc_z_max = 10.1, T-3 second acc_z_max = 10.5, T-4 second acc_z_max = 10.2. The filtered current value = (12.5 + 10.3 + 10.1 + 10.5 + 10.2) / 5 = 10.72. Although this filtered value 10.72 is still higher than the normal value, it smooths the single peak value of 12.5, making the subsequent anomaly detection algorithm more focused on the continuous jolt section rather than the isolated single impact that may not cause damage. After the above steps, a high-quality, analyzable sensor data stream is obtained, which is a time-ordered, time-aligned and filtered multi-dimensional data point sequence.

[0082] In step S202, the sensor data stream is processed to identify and divide a plurality of continuous dynamic driving scene segments.

[0083] Exemplarily, the cloud platform divides the sensor data stream into a plurality of clustering analysis windows with a non-overlapping time length of 30 seconds, calculates a feature vector of each window (including window acceleration standard deviation, window speed change rate, and window tilt angle fluctuation rate), clusters the feature vectors into four typical driving scenes (smooth high-speed, urban road conditions, bumpy road sections, and continuous curves / ramps) using a K-means clustering algorithm, and assigns cluster labels, then merges the windows with the same label and adjacent intervals less than 5 seconds, and finally converts the original data stream into a series of dynamic driving scene segments with different time lengths and explicit business meanings.

[0084] In one possible implementation, the sensor data stream is processed to identify and divide a plurality of continuous dynamic driving scene segment packages, specifically including: dividing the sensor data stream into a plurality of clustering analysis windows according to a preset time length; calculating a window feature vector of each clustering analysis window, the window feature vector including window acceleration standard deviation, window speed change rate, and window tilt angle fluctuation rate; processing the window feature vector through a clustering analysis algorithm to obtain a cluster label, and assigning the cluster label to the clustering analysis window corresponding to the window feature vector; merging adjacent clustering analysis windows with the same cluster label and adjacent time intervals less than a preset threshold to generate a plurality of continuous dynamic driving scene segments.

[0085] Specifically, the sensor data stream is divided into a plurality of clustering analysis windows, for example, the preset time length of each clustering analysis window is set to 30 seconds. These continuous sensor data streams are non-overlappingly cut into continuous 30-second windows in chronological order, for example, window 1 (W1) contains 30 data points from T0 to T0+29 seconds, window 2 (W2) contains 30 data points from T0+30 seconds to T0+59 seconds, and so on, to divide into a plurality of clustering analysis windows.

[0086] Further, a window feature vector is calculated for each cluster analysis window. For example, for each 30-second window divided, a window feature vector is calculated, which quantitatively describes the core driving features in the 30 seconds with a set of numerical values. The vector contains the following three key indicators: window acceleration standard deviation (Wacc_stddev), window velocity change rate (Wvel_change_rate), and window roll angle volatility (Wroll_volatility). Among them, Wacc_stddev is obtained by extracting the AccZMax_Filtered sequence of 30 data points in the window, and then calculating the standard deviation of the sequence; Wvel_change_rate is obtained by dividing the difference between the speed at the end of the window and the speed at the beginning of the window by the window length (30 seconds); Wroll_volatility is obtained by extracting the 30 roll angle data in the window, calculating the standard deviation, and the roll angle data is also collected by the MCU in the vehicle terminal. For example, three consecutive windows are analyzed, and the feature vectors obtained are as follows: W1: {Wacc_stddev: 0.15, Wvel_change_rate: 0.1, Wroll_volatility: 0.2}, indicating that the road section has little bump, the speed is basically unchanged, and the vehicle body is stable. W2: {Wacc_stddev: 2.85, Wvel_change_rate: -0.5, Wroll_volatility: 0.8}, indicating that the road section has very severe bump, the speed slightly decreases, and the vehicle body has certain shaking. W3: {Wacc_stddev: 3.10, Wvel_change_rate: -0.2, Wroll_volatility: 0.9}, indicating that the road section has continuous and slightly intensified bump, the speed is basically stable at low speed, and the vehicle body shaking is similar to that of W2 road section.

[0087] Further, the window feature vectors are processed by a clustering algorithm to obtain and assign cluster labels. For example, all window feature vectors are taken as input, and a clustering algorithm (such as K-means clustering algorithm) is used to group the vectors that are close in "distance" in the feature vector space into the same "cluster". For example, the feature vectors of W2 and W3 are very close in the three-dimensional space, while they are far away from the vector of W1. The number of clusters to be formed is pre-set (such as K = 4), representing four typical driving scenarios. After clustering, each cluster is assigned a numerical ID as a cluster label. For example: cluster 0 represents a smooth highway scenario (characteristics: low bumpiness, low speed variation, low jolt); cluster 1 represents an urban road scenario (characteristics: low bumpiness, high speed variation, medium jolt); cluster 2 represents a bumpy road scenario (characteristics: high bumpiness, low speed variation, medium-high jolt); and cluster 3 represents a continuous curve / ramp scenario (characteristics: medium bumpiness, medium speed variation, high jolt). The cluster label of each window feature vector is assigned to the corresponding clustering window. For example, the feature vector of W1 falls into "cluster 0", so W1 is labeled 0; the feature vector of W2 falls into "cluster 2", so W2 is labeled 2; and the feature vector of W3 also falls into "cluster 2", so W3 is also labeled 2.

[0088] Further, windows with the same label and in succession are merged into a longer, practically meaningful driving scenario segment. All windows with cluster labels are scanned in chronological order, and windows with the same cluster label and adjacent time intervals less than a pre-set threshold (such as 5 seconds) are checked. For example, if W(n) and W(n+1) have the same label, they are merged. This merging process continues until a window W(n+m) with a different cluster label is encountered. For example, consecutive windows with cluster label 0 (such as 10 windows) are merged into a "smooth highway" scenario segment, which has a duration of 10*30 = 300 seconds (i.e. 5 minutes). After the above steps, the cloud platform converts the original sensor data into a series of dynamic driving scenario segments with clear business meaning and different durations. For example: segment 1: {start time: T1, end time: T2, duration: 300 seconds, scenario type: smooth highway}; segment 2: {start time: T2, end time: T3, duration: 90 seconds, scenario type: bumpy road}; and segment 3: {start time: T3, end time: T4, duration: 180 seconds, scenario type: urban road}.

[0089] In step S203, the target dynamic driving scenario segment at the current time is determined, a pre-set number of data points from the start time of the target dynamic driving scenario segment are selected as modeling samples, and a multi-dimensional baseline feature statistical model is constructed for the target dynamic driving scenario segment based on the modeling samples.

[0090] Exemplarily, the cloud platform determines a target dynamic driving scene segment at the current time (e.g., a bump road segment segment with ID Segment_ID_001), selects continuous data points in the 60 seconds after the start of the segment as modeling samples, calculates the acceleration standard deviation, the speed change rate, and the tilt angle volatility for each data point to form 60 three-dimensional feature vectors, and then calculates the mean vector μ (e.g., {2.6, -0.35, 0.75}) and the covariance matrix Σ (containing the variances of the features and the covariances between the features) of the feature vector set through the multi-dimensional Gaussian distribution, to finally construct a multi-dimensional baseline feature statistical model N(μ, Σ) specific to the segment.

[0091] In a possible implementation, a target dynamic driving scene segment at the current time is determined, a preset number of continuous data points starting from the start time of the target dynamic driving scene segment are selected as modeling samples, and a multi-dimensional baseline feature statistical model is constructed for the target dynamic driving scene segment based on the modeling samples. Specifically, a multi-dimensional feature corresponding to a second data point in the modeling samples is calculated to form a feature vector of the second data point, the second data point being any data point in the modeling samples, and the feature vector including an acceleration standard deviation, a speed change rate, and a tilt angle volatility. All the feature vectors in the modeling samples are combined into a feature vector set. A statistical distribution parameter of the distribution of the feature vector set in a multi-dimensional feature space is calculated. Based on the statistical distribution parameter, the multi-dimensional baseline feature statistical model is constructed.

[0092] Specifically, the target segment is determined and the modeling samples are selected. First, a target dynamic driving scene segment at the current time is determined, for example, at t0=14:30:00, a scene switching event is detected, that is, the "target dynamic driving scene segment at the current time" is determined to be a newly started dynamic driving scene segment, and is assigned a unique ID, for example, Segment_ID_001 (type: bump road segment). Then, a preset number of continuous data points starting from the start time of the segment are selected as modeling samples. For example, the preset number is 60, and since the sensor data stream is regularized by seconds, this corresponds to the data in the first minute (from 14:30:30 to 14:30:59) after the start of the segment. Then, each data point in the 60-second modeling sample is characterized to quantify its instantaneous state. For example, around each data point, the acceleration standard deviation (acc_stddev), the speed change rate (vel_change_rate), and the tilt angle volatility (roll_volatility) are calculated, and the calculation method is the same as the foregoing method, which is not described herein again. Finally, a set of 60 three-dimensional feature vectors is obtained. This set can be represented as a 60x3 matrix in mathematics.

[0093] Further, the statistical distribution parameters of the feature vector set are calculated and a model is constructed. For example, the mean vector μ and the covariance matrix Σ of the feature vector set are calculated respectively by a Multivariate Gaussian Distribution. Wherein, the mean vector μ is calculated: the average value of each column (each feature dimension) in the feature vector set is calculated respectively, such as μ = {2.6, -0.35, 0.75}, this mean vector μ represents the most typical and average state of driving on the Segment_ID_001 segment. That is: the average bump degree is 2.6, the average is in a slight deceleration state (-0.35), and the average vehicle body shaking is 0.75. This is the behavior center of the segment. The covariance matrix Σ is calculated: the diagonal elements are the variances of the three features respectively, and the larger the variance, the wider the numerical fluctuation range of the feature. For example, the variance of acc_stddev is large, indicating that the bump degree of the road is high and low, and changes greatly. The non-diagonal elements are the covariances between different features, indicating the correlation between the features. For example, if the covariance of vel_change_rate and roll_volatility is positive, it means that acceleration or deceleration usually accompanies more severe vehicle body shaking on the road segment. For example,

[0094]

[0095] Wherein, Σ[1, 1] = 0.8 (the variance of acc_stddev) is the largest, indicating that the severity of the bump itself is very unstable on the bump road segment; Σ[2, 2] = 0.2 (the variance of vel_change_rate) is the smallest, indicating that the acceleration and deceleration behavior of the vehicle is relatively the most stable on the road; Σ[3, 3] = 0.3 (the variance of roll_volatility) is between the two, indicating that the amplitude of vehicle shaking has certain fluctuation. Based on the calculated mean vector μ and covariance matrix Σ, a multivariate baseline feature statistical model specific to Segment_ID_001 segment is constructed. This model is a multivariate Gaussian distribution N(μ, Σ) completely defined by μ and Σ.

[0096] Step S204, input the first data point of the current time to the multivariate baseline feature statistical model, and calculate the abnormal score of the first data point.

[0097] Exemplarily, the cloud platform extracts the multivariate features of the current time data point, generates the target feature vector, inputs it into the corresponding multivariate baseline model to calculate the Mahalanobis distance, converts the Mahalanobis distance into a probability conforming to the normal baseline, generates an abnormal value according to "the lower the probability, the higher the abnormal value", and obtains an abnormal score after the abnormal value is normalized by nonlinear mapping.

[0098] In a possible implementation, the first data point at the current time is input into the multi-dimensional baseline feature statistical model, and an anomaly score of the first data point is calculated, specifically comprising: extracting multi-dimensional features of the first data point at the current time to generate a target feature vector; inputting the target feature vector into the multi-dimensional baseline feature statistical model, and calculating a probability that the target feature vector conforms to a normal driving behavior baseline defined by the multi-dimensional baseline feature statistical model by using a statistical distribution parameter; generating an anomaly value for the first data point according to the probability, wherein the lower the probability, the higher the anomaly value; and performing normalization processing on the anomaly value to obtain an anomaly score.

[0099] Specifically, multi-dimensional features of the data point at the current time are extracted to generate a target feature vector. For example, the cloud platform receives a data point at t1=14:31:15. At t1, the target vehicle is still driving on the previously identified Segment_ID_001 (bumpy road segment). At this time, the right front wheel of the vehicle runs over a deep pit that has not been discovered in time, resulting in a sharp impact far exceeding the average bump level of the road segment. By using the same method as in the foregoing modeling, a 5-second sliding window (from 14:31:13 to 14:31:17) centered at t1 is used to calculate the multi-dimensional features of the data point. Since the vehicle runs over the deep pit, the data in the 5-second window will show significant anomalies. Acceleration standard deviation (acc_stddev): the impact value of the Z-axis will have a very abrupt peak, causing the standard deviation of the AccZMax_Filtered values of the 5 data points to increase sharply. Velocity change rate (vel_change_rate): the driver may subconsciously step on the brake hard, causing the speed to drop rapidly in a short time. Roll volatility (roll_volatility): unilateral wheel sinking will cause the vehicle body to tilt and rebound sharply, causing the standard deviation of the roll angle readings to also increase significantly. Calculation result example: the system calculates the target feature vector of the data point at t1=14:31:15 as V_target={8.5, -2.1, 3.2}. Compared with the center of the established multi-dimensional baseline feature statistical model (i.e., the "normal" baseline): μ={2.6, -0.35, 0.75}. Obviously, V_target deviates from the "normal" mean value in each dimension.

[0100] Further, the probability that the target feature vector conforms to the normal baseline is calculated. For example, the target feature vector V_target={8.5, -2.1, 3.2} is input into the multi-dimensional baseline feature statistical model N(μ,Σ) established for the current target dynamic driving scene segment. The Mahalanobis distance of V_target is calculated by using the two parameters μ and Σ, and the calculation formula is as follows:

[0101] D(V_target) 2= (V_target - μ) T ∑ -1 (V_target - μ)

[0102] where D(V_target) denotes Mahalanobis distance, V_target denotes target feature vector, μ denotes mean vector, ∑ denotes covariance matrix, and T denotes transpose. The calculated D(V_target) 2 ≈ 85.17. This number is much larger than 1. In a multi-dimensional Gaussian distribution, if a point is “normal”, the square of its Mahalanobis distance is usually in the single digit. 85.17 is a very large value. Statistically, the farther a point is from the center of a Gaussian distribution (the larger the Mahalanobis distance), the less likely it is to belong to this distribution. The cloud platform converts this Mahalanobis distance into a specific probability value through the probability density function. The probability that V_target is consistent with the normal baseline is P = 0.00001. This is a very low probability value.

[0103] Further, the abnormal value is generated according to the probability, and the abnormal score is obtained by normalizing the abnormal value. For example, the abnormal value is inversely proportional to the probability. The lower the probability, the more rare and abnormal the event, and therefore the higher the corresponding abnormal value. The abnormal value can be obtained by taking the reciprocal of the probability. For example, the abnormal value Anomaly Value = 1 / P = 1 / 0.00001 = 100000. The cloud platform uses a nonlinear mapping function (such as a Sigmoid function or a variant thereof) to normalize the abnormal value and map it to the interval of 0-100. The mapping logic is: when the original abnormal value is very small (i.e., the probability is very high and the behavior is very normal), the normalized score is close to 0. When the original abnormal value is very large (i.e., the probability is very low and the behavior is very abnormal), the normalized score will quickly approach 100. This mapping relationship makes the score very sensitive in the critical region of “normal” and “abnormal”, and tends to be stable in the two extreme regions. When the abnormal value Anomaly Value = 100000 is input into the normalization function, the final abnormal score of the data point at time t1 = 14:31:15 is 98.

[0104] In step S205, when the abnormal score is greater than the preset risk alarm score value, a risk event record is generated based on the first data point and the abnormal score.

[0105] For example, when the abnormal score exceeds the preset alarm score value, the platform locks the current data point and the target feature vector, analyzes the location information, maps the score interval to the preset risk level, determines the main abnormal dimension by comparing the mean vector, matches the risk type, and finally generates a complete risk event record by combining the information.

[0106] In a possible implementation, when the anomaly score is greater than a preset risk alarm score, a risk event record is generated based on the first data point and the anomaly score, specifically including: when the anomaly score exceeds the preset risk alarm score, locking the first data point and the corresponding target feature vector; extracting target location information from the first data point; mapping the anomaly score to a preset risk level based on the numerical interval in which the anomaly score is located; analyzing each dimension value of the target feature vector, calculating the contribution weight of each dimension to the anomaly score, and determining the dimension with the highest contribution weight as the main abnormal dimension; matching a preset risk type mapping table according to the main abnormal dimension and the corresponding numerical deviation degree to determine the risk type, wherein the numerical deviation degree is the deviation amplitude of the numerical value of the main abnormal dimension from the normal driving behavior baseline; combining the target location information, the risk level, and the risk type to generate the risk event record.

[0107] Specifically, the calculated anomaly score is 98, and the platform preset risk alarm score is 85. At this time, the anomaly score is greater than the risk alarm score, and the risk event record generation program is triggered. At this time, all original information of the data point at t1 = 14:31:15 is snapshoted and locked to prevent being overwritten by subsequent data streams. At the same time, the corresponding target feature vector V_target = {8.5, -2.1, 3.2} is also locked as the core analysis object of this event. From the locked data point, the target location information recorded by the GPS module is directly analyzed, for example, the longitude is 113.587621 and the latitude is 23.459873.

[0108] Further, in order to intuitively understand the severity of the event, the anomaly score of 0-100 is mapped to a preset risk level. For example, the preset risk level mapping table is as follows:

[0109] Abnormal score interval Risk level Suggested action 0-60 Low risk Record, no alarm needed 61-85 Medium risk Yellow light alert, and keep an eye on it 86-100 High risk Red light alert, and issue an alarm

[0110] The current anomaly score is 98, which falls into the [86-100] interval, so the risk level of this event is determined to be high risk.

[0111] Further, three latitudes in the target feature vector are analyzed, and the values in each dimension are calculated. Compared with the "normal" baseline in the multi-dimensional baseline feature statistical model, how large is the relative amplitude of the deviation. For example, the difference between the target feature vector V_target = {8.5, -2.1, 3.2} and the mean vector μ = {2.6, -0.35, 0.75} in each dimension is: the deviation value of acc_stddev = |8.5-2.6| = 5.9, the deviation value of vel_change_rate = |-2.1-(-0.35)| = 4.15, and the deviation value of roll_volatility = |3.2-0.75| = 2.45. Among them, the deviation value of acc_stddev is the largest, so the main abnormal dimension is determined as the acceleration standard deviation acc_stddev. Based on the main abnormal dimension analyzed, the preset risk type mapping table is queried to determine the risk type of the event. The preset risk type mapping table is as follows:

[0112] Primary abnormality dimension Numerical deviation direction Risk type Acceleration standard deviation Far from baseline Severe vertical impact / jolt Rate of change of speed Far greater than baseline Abrupt acceleration Rate of change of speed Far less than baseline Abrupt deceleration Roll angle fluctuation rate Far from baseline Severe shaking / rollover risk

[0113] For example, according to the main abnormal dimension obtained in the foregoing as the acceleration standard deviation, and the value deviates far from the baseline, it is determined that the risk type of the event is a violent vertical impact / jolt. Finally, all the structured information analyzed in the foregoing is combined to generate a complete risk event record that can be archived and sent.

[0114] In step S206, a warning instruction is generated based on the risk event record, and the warning instruction is sent to the logistics monitoring center and the vehicle terminal of the target delivery vehicle.

[0115] For example, the cloud platform parses the location information, risk level and risk type from the risk event record, generates corresponding warning contents for the logistics monitoring center and the vehicle terminal, respectively, matches the warning form according to the risk level, and finally encapsulates the warning content and the warning form into a JSON instruction and pushes it to the corresponding terminal.

[0116] In one possible implementation, a warning instruction is generated based on the risk event record, and the warning instruction is sent to the logistics monitoring center and the vehicle terminal of the target delivery vehicle, specifically including: parsing the target location information, risk level and risk type from the risk event record; generating warning content according to the risk type and the target location information; selecting one or more warning forms from a preset warning strategy library according to the risk level, the warning forms including a sound alarm, a visual highlight prompt or a text message; encapsulating the warning content and the warning form into a warning instruction; and synchronously sending the warning instruction to the logistics monitoring center and the vehicle terminal of the target delivery vehicle.

[0117] Specifically, the target location information, risk level and risk type are parsed from the risk event record. For example, the target location information is parsed as: longitude 113.587621, latitude 23.459873; the risk level is parsed as: “high risk”; and the risk type is parsed as: “severe vertical impact / jolt”. According to the parsed information, corresponding warning contents are generated for the logistics monitoring center and the vehicle terminal, respectively. For example, for the logistics monitoring center, the risk type is combined with the location information, and the vehicle ID is added to form a complete descriptive text, such as “high risk warning: vehicle ‘truck_id_001’ encountered ‘severe vertical impact’ near G107 National Highway (longitude and latitude: 113.59, 23.46), and the goods may be damaged. Please check immediately!”; for the vehicle terminal, complex location and technical terms are discarded, and colloquial phrases are used for prompting, such as “warning: severe impact encountered! Please stop and check the safety of the goods later.”

[0118] Further, different warning forms are selected according to different risk levels. For example, this event is ‘high risk’, which needs to trigger the highest level of warning strategy. For example, according to the risk level, the corresponding warning form combination is matched from the strategy library. For the logistics monitoring center, visual high-lighting prompt is performed: on the electronic map, the icon representing the target delivery vehicle truck_id_001 is changed to a red high-frequency flashing state, and the map view is automatically centered to this point. For the vehicle terminal, sound warning is performed: through the vehicle speaker, a warning audio is played at the maximum volume.

[0119] Further, the warning content and the warning form are packaged as a warning instruction and sent. For example, structured warning instructions (such as JSON data packets) are generated for the logistics monitoring center and the vehicle terminal, respectively. The two warning instructions are respectively pushed to the front-end application of the logistics monitoring center and the vehicle terminal of the target delivery vehicle through their respective communication channels.

[0120] Figure 3 is a module schematic diagram of a logistics cloud platform information transmission supervision system based on big data in the embodiments of the present application, as shown in Figure 3 The system comprises a data acquisition module 301, a segment division module 302, a model construction module 303, a score calculation module 304, a risk record module 305 and a warning generation module 306, wherein:

[0121] The data acquisition module 301 is configured to continuously receive the sensor data stream uploaded by the vehicle terminal of the target delivery vehicle, and the sensor data stream comprises a plurality of data points;

[0122] The segment division module 302 is configured to process the sensor data stream, identify and divide a plurality of continuous dynamic driving scene segments;

[0123] The model construction module 303 is configured to determine a target dynamic driving scene segment at a current time, select a preset number of data points starting from a starting time in the target dynamic driving scene segment as modeling samples, and construct a multi-dimensional baseline feature statistical model for the target dynamic driving scene segment based on the modeling samples.

[0124] The score calculation module 304 is configured to input the first data point at the current time into the multi-dimensional baseline feature statistical model, and calculate an abnormal score of the first data point.

[0125] The risk record module 305 is configured to, when the abnormal score is greater than a preset risk alarm score, generate a risk event record based on the first data point and the abnormal score.

[0126] The early warning generation module 306 is configured to generate a pre-warning instruction based on the risk event record, and send the pre-warning instruction to a logistics monitoring center and a vehicle-mounted terminal of the target delivery vehicle.

[0127] Optionally, the data acquisition module 301 is further configured to:

[0128] receive an original data packet uploaded by the target delivery vehicle, the original data packet containing three-axis acceleration, instantaneous speed and position information; parse the original data packet, and perform time sequence alignment on the three-axis acceleration, instantaneous speed and position information according to a time stamp to obtain multi-dimensional time sequence data;

[0129] filter the multi-dimensional time sequence to obtain a sensor data stream.

[0130] Optionally, the segment division module 302 is further configured to:

[0131] divide the sensor data stream into a plurality of clustering analysis windows according to a preset time length;

[0132] calculate a window feature vector of each clustering analysis window, the window feature vector including a window acceleration standard deviation, a window speed change rate and a window tilt angle fluctuation rate;

[0133] process the window feature vector through a clustering analysis algorithm to obtain a cluster label, and assign the cluster label to the clustering analysis window corresponding to the window feature vector;

[0134] merge adjacent clustering analysis windows with the same cluster label and a time interval less than a preset threshold to generate a plurality of continuous dynamic driving scene segments.

[0135] Optionally, the model construction module 303 is further configured to:

[0136] corresponding to the second data point in the modeling sample, and form a feature vector of the second data point, the second data point being any data point in the modeling sample, the feature vector including an acceleration standard deviation, a speed change rate, and a tilt angle fluctuation rate;

[0137] all the feature vectors in the modeling sample are combined into a feature vector set;

[0138] a statistical distribution parameter of the distribution rule of the feature vector set in the multi-dimensional feature space is calculated;

[0139] a multi-dimensional baseline feature statistical model is constructed based on the statistical distribution parameter.

[0140] Optionally, the score calculation module 304 is further configured to:

[0141] multi-dimensional features of the first data point at the current time are extracted to generate a target feature vector;

[0142] the target feature vector is input into the multi-dimensional baseline feature statistical model, and a probability that the target feature vector conforms to a normal driving behavior baseline defined by the multi-dimensional baseline feature statistical model is calculated using the statistical distribution parameter;

[0143] an abnormal value is generated for the first data point according to the probability, wherein the lower the probability, the higher the abnormal value;

[0144] the abnormal value is normalized to obtain an abnormal score.

[0145] Optionally, the risk record module 305 is further configured to:

[0146] when the abnormal score exceeds a preset risk warning score value, the first data point and the corresponding target feature vector are locked;

[0147] target location information is extracted from the first data point;

[0148] the abnormal score is mapped to a preset risk level based on the numerical interval in which the abnormal score is located;

[0149] each dimension value of the target feature vector is analyzed, a contribution weight of each dimension to the abnormal score is calculated, and the dimension with the highest contribution weight is determined as a main abnormal dimension;

[0150] a preset risk type mapping table is matched according to the main abnormal dimension and a numerical deviation degree of the main abnormal dimension, to determine a risk type, wherein the numerical deviation degree is a deviation amplitude of the numerical value of the main abnormal dimension from the normal driving behavior baseline;

[0151] the target location information, the risk level, and the risk type are combined to generate a risk event record.

[0152] Optionally, the early warning generation module 306 is further configured to:

[0153] resolve the target location information, the risk level and the risk type from the risk event record;

[0154] generate the early warning content according to the risk type and the target location information;

[0155] select one or more early warning forms from a preset early warning strategy library according to the risk level, the early warning forms including a sound alarm, a visual highlight prompt or a text message;

[0156] encapsulate the early warning content and the early warning form into an early warning instruction;

[0157] synchronously send the early warning instruction to a vehicle-mounted terminal corresponding to the target delivery vehicle in the logistics monitoring center.

[0158] It should be noted that the system provided in the above embodiments is only used as an example to divide the above functional modules when realizing its functions. In actual applications, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be described here.

[0159] The embodiment also discloses an electronic device, which refers to Figure 4 The electronic device can include at least one processor 401, at least one communication bus 402, a user interface 403, a network interface 404, and at least one memory 405.

[0160] The communication bus 402 is used to realize the connection and communication between the components.

[0161] The user interface 403 can include a display screen (Display) and a camera (Camera), and the optional user interface 403 can further include a standard wired interface and a wireless interface.

[0162] The network interface 404 can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0163] The processor 401 can include one or more processing cores. The processor 401 connects various parts within the server through various interfaces and lines, performs various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 405, and calling data stored in the memory 405. Alternatively, the processor 401 can be implemented in at least one of a hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 401 can integrate a combination of one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. Among them, the CPU mainly processes operating systems, user interfaces, and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; and the modem is used for processing wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 401, but can be realized by a separate chip.

[0164] The memory 405 can include a random access memory (RAM) and a read-only memory (ROM). Alternatively, the memory 405 includes a non-transitory computer-readable storage medium. The memory 405 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 405 can include a program storage area and a data storage area, wherein the program storage area can store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area can store data involved in the above-mentioned various method embodiments, etc. The memory 405 can alternatively be at least one storage device located away from the aforementioned processor 401. As shown, the memory 405 as a computer storage medium can include an operating system, a network communication module, a user interface module, and an application program of a logistics cloud platform information transmission monitoring method based on big data. Figure 4

[0165] In Figure 4 ​In the electronic device shown, the user interface 403 is mainly used to provide an interface for user input, and obtain data input by the user; and the processor 401 can be used to call an application program of a big data-based logistics cloud platform information transmission monitoring method stored in the memory 405, and when executed by one or more processors 401, the electronic device executes the method of one or more of the above-described embodiments.

[0166] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all described as a series of action combinations, but those skilled in the art should know that the present application is not limited to the order of the actions described, because according to the present application, certain steps can be performed in other order or at the same time. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0167] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0168] In several embodiments provided by the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic. The division of the units is only a logical function division. There can be another division manner for actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different parts can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical or other forms.

[0169] The units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0170] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically independently, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0171] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable memory. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a memory 405 and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned memory 405 includes: a U disk, a mobile hard disk, a magnetic or optical disk and various media that can store program codes.

[0172] The above-described are only exemplary embodiments of the present disclosure, and cannot limit the scope of the present disclosure. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure are still within the scope of the present disclosure. Other embodiments of the present disclosure will be readily apparent to those skilled in the art upon considering the disclosure. The present application is intended to cover any variations, uses or adaptive changes of the present disclosure that follow the general principles of the present disclosure and include common knowledge or conventional technical means in the art that are not described in the present disclosure. The scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for monitoring information transmission on a logistics cloud platform based on big data, characterized in that, Applied to a server, the method includes: Continuously receive sensor data streams uploaded by the on-board terminal of the target carrier vehicle, wherein the sensor data streams contain multiple data points; The sensor data stream is processed to identify and segment multiple consecutive dynamic driving scenario segments; Determine the target dynamic driving scenario segment at the current moment, select a preset number of data points from the start time of the target dynamic driving scenario segment as modeling samples, and construct a multidimensional baseline feature statistical model for the target dynamic driving scenario segment based on the modeling samples; Input the first data point at the current moment into the multidimensional baseline feature statistical model to calculate the anomaly score of the first data point; When the anomaly score is greater than the preset risk alarm score, a risk event record is generated based on the first data point and the anomaly score. Based on the risk event records, an early warning instruction is generated and sent to the logistics monitoring center and the vehicle terminal of the target transport vehicle.

2. The method according to claim 1, characterized in that, The continuous reception of sensor data streams uploaded by the onboard terminal of the target transport vehicle includes: Receive the raw data packet uploaded by the target carrier vehicle, the raw data packet containing triaxial acceleration, instantaneous velocity and position information; The original data packet is parsed, and the triaxial acceleration, instantaneous velocity, and position information are time-series aligned according to the timestamp to obtain multidimensional time series data; The multidimensional time series is filtered to obtain the sensor data stream.

3. The method according to claim 1, characterized in that, The process of processing the sensor data stream to identify and segment multiple consecutive dynamic driving scenario segments includes: The sensor data stream is divided into multiple clustering analysis windows according to a preset duration; Calculate the window feature vector for each cluster analysis window, wherein the window feature vector includes the standard deviation of window acceleration, the rate of change of window velocity, and the fluctuation rate of window tilt angle; The feature vector of the window is processed by a clustering analysis algorithm to obtain cluster labels, and the cluster labels are assigned to the clustering analysis window corresponding to the feature vector of the window. Adjacent cluster analysis windows with the same cluster label and an adjacent time interval of less than a preset threshold are merged to generate multiple consecutive dynamic driving scenario segments.

4. The method according to claim 1, characterized in that, The step of determining the target dynamic driving scenario segment at the current moment, selecting a predetermined number of consecutive data points from the start time of the target dynamic driving scenario segment as modeling samples, and constructing a multidimensional baseline feature statistical model for the target dynamic driving scenario segment based on the modeling samples includes: Calculate the multidimensional features corresponding to the second data point in the modeling sample to form the feature vector of the second data point. The second data point is any data point in the modeling sample. The feature vector includes the standard deviation of acceleration, the rate of change of velocity, and the rate of fluctuation of tilt angle. Combine all the feature vectors in the modeling sample into a feature vector set; Calculate the statistical distribution parameters of the distribution pattern of the feature vector set in the multidimensional feature space; Based on the statistical distribution parameters, the multidimensional baseline feature statistical model is constructed.

5. The method according to claim 4, characterized in that, The step of inputting the first data point at the current moment into the multidimensional baseline feature statistical model and calculating the anomaly score of the first data point includes: Extract the multidimensional features of the first data point at the current time and generate the target feature vector; The target feature vector is input into the multidimensional baseline feature statistical model, and the probability that the target feature vector conforms to the normal driving behavior baseline defined by the multidimensional baseline feature statistical model is calculated using the statistical distribution parameters. Based on the probability, outliers are generated for the first data point, wherein the lower the probability, the higher the outlier. The outliers are normalized to obtain the outlier score.

6. The method according to claim 5, characterized in that, When the anomaly score is greater than a preset risk alarm score, the generation of a risk event record based on the first data point and the anomaly score includes: When the abnormal score exceeds the preset risk alarm score, the first data point and the corresponding target feature vector are locked. Extract the target location information from the first data point; Based on the numerical range of the abnormal score, the abnormal score is mapped to a preset risk level; Analyze the values ​​of each dimension of the target feature vector, calculate the contribution weight of each dimension to the anomaly score, and determine the dimension with the highest contribution weight as the main anomaly dimension; Based on the main anomaly dimensions and their corresponding numerical deviations, a preset risk type mapping table is matched to determine the risk type, wherein the numerical deviation is the deviation of the value of the main anomaly dimension from the baseline of normal driving behavior. The target location information, the risk level, and the risk type are combined to generate the risk event record.

7. The method according to claim 6, characterized in that, The step of generating an early warning instruction based on the risk event record and sending the early warning instruction to the logistics monitoring center and the vehicle-mounted terminal of the target carrier includes: The target location information, the risk level, and the risk type are parsed from the risk event records; Based on the risk type and the target location information, generate warning content; Based on the risk level, select one or more warning forms from a preset warning strategy library, including sound alarms, visual highlighting prompts, or text messages; The warning content and the warning format are encapsulated into a warning instruction; The warning command is simultaneously sent to the corresponding vehicle terminal of the logistics monitoring center and the target transport vehicle.

8. A logistics cloud platform information transmission and monitoring system based on big data, characterized in that, The system includes a data acquisition module, a segmentation module, a model building module, a score calculation module, a risk recording module, and an early warning generation module, among which: The data acquisition module is configured to continuously receive sensor data streams uploaded by the on-board terminal of the target transport vehicle, and the sensor data streams contain multiple data points. The segmentation module is configured to process the sensor data stream and identify and segment multiple consecutive dynamic driving scenario segments. The model building module is configured to determine the target dynamic driving scenario segment at the current moment, select a preset number of data points from the start time of the target dynamic driving scenario segment as modeling samples, and build a multidimensional baseline feature statistical model for the target dynamic driving scenario segment based on the modeling samples. The scoring calculation module is configured to input the first data point at the current moment into the multidimensional baseline feature statistical model to calculate the abnormal score of the first data point. The risk recording module is configured to generate a risk event record based on the first data point and the abnormal score when the abnormal score is greater than a preset risk alarm score. The early warning generation module is configured to generate early warning instructions based on the risk event records and send the early warning instructions to the logistics monitoring center and the vehicle-mounted terminal of the target carrier vehicle.

9. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions. The user interface and the network interface are both used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1-7.

Images