Fresh food cold chain tracing and early warning system based on internet of things technology

Abstract

The application discloses a fresh food cold chain tracing and early warning system based on Internet of Things technology, relates to the technical field of fresh food cold chain tracing, integrates sensing modules such as GPS positioning, temperature and humidity monitoring, electronic fence management, parking identification, cargo compartment state monitoring, comprehensively judges the actual stop behavior of a vehicle through a data fusion analysis module, generates a credibility score, a path execution judgment module further intelligently compares the score result with a preset path of a distribution task, accurately judges whether each stop task is completed, if it is identified that the task is not completed, an abnormality early warning module timely reports abnormal information to a dispatch center, the whole process intelligent identification and real-time early warning of the distribution behavior are realized, and the accuracy, completeness and system response capability of cold chain tracing are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of fresh food cold chain traceability technology, specifically to a fresh food cold chain traceability and early warning system based on Internet of Things (IoT) technology. Background Technology

[0002] The fresh food cold chain traceability and early warning system based on Internet of Things (IoT) technology refers to the use of IoT technologies such as sensors, wireless communication, and cloud computing to collect and monitor key data such as temperature, humidity, and location of fresh products in real time throughout the entire cold chain transportation and storage process. This enables comprehensive traceability of the source, circulation path, and quality status of fresh products, and timely issuance of early warnings when abnormalities are detected (such as excessive temperature), thereby ensuring food safety and improving logistics efficiency.

[0003] The existing technology has the following shortcomings:

[0004] In urban cold chain delivery, when multiple stop points are densely located (e.g., less than 500 meters apart), if the platform system's GPS sampling frequency is low and only uses entering the electronic fence as the criterion for determining a stop, the system may misjudge a vehicle's brief stop as a missed stop or skipped a stop if it is not accurately identified. This error can cause data loss in order tracking, incomplete user interface display, and even be mistakenly identified as delivery fraud during audits, thus misleading consumers or regulatory agencies into believing there is a risk of a cold chain disruption. Summary of the Invention

[0005] The purpose of this invention is to provide a fresh food cold chain traceability and early warning system based on Internet of Things (IoT) technology to address the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a fresh food cold chain traceability and early warning system based on Internet of Things technology, including a GPS positioning module, a temperature and humidity sensing module, an electronic fence management module, a parking identification module, a cargo compartment status monitoring module, a data fusion analysis module, a route execution determination module, and an anomaly early warning module;

[0007] The GPS positioning module is used to collect real-time location information of cold chain vehicles;

[0008] Temperature and humidity sensing module, used to collect temperature and humidity data of the refrigerated environment;

[0009] The electronic fence management module is used to set the geofence information for each stop point in the cold chain delivery task;

[0010] The parking recognition module is used to identify actual parking behavior based on changes in vehicle speed and the duration of continuous stillness.

[0011] The cargo compartment status monitoring module is used to detect the open or closed status of the vehicle-mounted cargo compartment;

[0012] The data fusion and analysis module is used to fuse and analyze GPS positioning information, parking recognition results and cargo compartment status data, and output the corresponding credibility score of the stop behavior;

[0013] The route execution determination module, based on the credibility score and the comparison results of the preset route and stop information in the delivery task, determines whether the vehicle has completed the stop task at the designated delivery point.

[0014] If the anomaly warning module fails to complete, it will generate corresponding warning information and send it to the dispatch center.

[0015] Preferably, the electronic fence management module supports custom polygon settings for fence areas and configures task numbers, time windows, and docking behavior requirements for each fence.

[0016] Preferably, the cargo compartment status monitoring module collects the cargo compartment opening and closing status based on the door magnetic sensor, and allows door opening operation when the vehicle is within the designated electronic fence. If door opening occurs outside the preset area, an abnormal warning is automatically triggered.

[0017] Preferably, the data fusion and analysis module specifically includes:

[0018] Acquire vehicle GPS trajectory information, parking recognition data, and cargo compartment status data;

[0019] The data is aligned according to timestamps;

[0020] Extract stop behavior feature values, including GPS dwell time deviation value, abnormal gate opening frequency value, and offset distance between the parking point and the center of the fence;

[0021] The feature values ​​are normalized, and the credibility score of the stopover behavior is calculated according to the weighted average formula, with the score range from 0 to 1.

[0022] Preferably, the method for extracting the GPS dwell time deviation value is as follows: Input GPS trajectory point sequence P = {(lat1,lon1,t1),(lat2,lon2,t2),…, ... n ,lon n ,t n )};lat i Let lon be the latitude of the i-th GPS point. i Let t be the longitude of the i-th GPS point. i Let be the timestamp of the i-th GPS point; map the GPS track points to actual roads or geographic areas, use map matching algorithms to correct GPS drift and jump points, and output a set of track points with matching results, marking whether they are within the target fence: P′={(lat i ,loni ,t i ,inzone)};inzone indicates that it is within the target fence; traverse the matching trajectory points, find the data segment with continuous in zone = true, get the actual entry time: Tenter, the actual exit time: Texit, calculate the actual dwell time Tactual: Tactual = Texit - Tenter; get the expected dwell time Texpected, calculate the GPS dwell time deviation value ΔT, the expression is: ΔT = Tactual - Texpected.

[0023] Preferably, the method for extracting outliers in the door opening frequency is as follows: each time the cargo compartment door is opened, the sensor records one event: D = {(t1, loc1), (t2, loc2), ..., (t m ,loc m Each data entry contains a door opening timestamp t. i Door opening position coordinates loc i m is the total number of coordinates; the preset tasks include: a list of stop points S = {s1, s2, ..., s...} z}; Whether each station requires the door to be opened; Set the expected number of openings as: Nexpected = z; That is: if there are z delivery points, the expected number of openings is z times, and count the actual number of openings: Nactual = len(D); Count the total number of actual opening events of the cargo compartment door, and define the abnormal value of the opening frequency ΔN as: ΔN = Nactual - Nexpected.

[0024] Preferably, the method for obtaining the offset distance is as follows: In the task settings, a fence is set for each delivery point, and the coordinates of its center point are: C = (lat c ,lon c ); that is, the center of the electronic fence of the station, which is preset as the ideal parking location; obtain the parking point coordinates of the vehicle from the parking recognition module: P = (lat p ,lon p The calculation is performed using the Earth-plane approximation method. Where: Δlat=lat p -lat c ; Δlon = lon p -lon c ; D is the offset distance.

[0025] Preferably, the credibility score of the stopover behavior is compared with the gradient standard threshold, which includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. The credibility score of the stopover behavior is compared with the first standard threshold and the second standard threshold respectively.

[0026] If the credibility score of the stopover behavior is greater than the second standard threshold, it is judged as a high-credibility delivery behavior; if the credibility score of the stopover behavior is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it is judged as a medium-credibility delivery behavior; if the credibility score of the stopover behavior is less than the first standard threshold, it is judged as a low-credibility delivery behavior.

[0027] Preferably, the preset route and stop information in the delivery task are compared, and the comparison results are combined with a reliability score for comprehensive analysis to determine whether the vehicle has completed the stop task at the designated delivery point. Specifically:

[0028] Preset path data R = {r1, r2, ..., r p}: p is the total number of path data in the sequence; the actual trajectory data P = {(lat1,lon1,t1),(lat2,lon2,t2),…, ... n ,lon n ,t n )};The credibility score for each stop point is Gi∈[0,1];Map matching is performed on the trajectory sequence to correct GPS jump points;

[0029] Calculate the trajectory similarity between the actual trajectory and the desired path segment: D dtw The actual DTW distance between the actual and preset trajectories; Sim i ∈[0,1]: Trajectory matching similarity;

[0030] Sim, the trajectory comparison results i A comprehensive analysis is performed with the behavioral credibility score Ci to define the final path execution credibility F. i :F i =α·Ci+(1-α)·Sim i ; α∈[0.5,0.9].

[0031] Preferably, the final path execution reliability is compared with a predetermined threshold. If the path execution reliability is greater than or equal to the predetermined threshold, the path execution is determined to be successful, indicating that the behavior of the vehicle at all stops during the entire delivery task conforms to the reliability rules set by the system.

[0032] If the reliability of the route execution is less than the predetermined threshold, it is determined to be an abnormal route execution, indicating that the vehicle has obvious abnormal behavior in this delivery task, and an abnormality warning is automatically sent to the dispatch platform.

[0033] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0034] 1. This invention overcomes the problem of misjudgment in stop identification caused by low-granularity GPS data and a single judgment mechanism in traditional cold chain delivery systems by introducing a multi-module collaborative Internet of Things (IoT) structure. By combining multi-dimensional sensing data such as GPS positioning, temperature and humidity sensing, fence recognition, parking detection, and cargo compartment opening and closing status, and outputting a stop behavior credibility score through a fusion analysis model, it not only significantly improves the accuracy and robustness of delivery behavior identification under high-density delivery stations, but also avoids common misjudgments such as missed judgments and skipping stations, thereby ensuring the integrity and verifiability of order traceability data.

[0035] 2. The path execution determination and hierarchical judgment mechanism proposed in this invention can intelligently distinguish the credibility level of delivery behavior. Combined with the anomaly warning module, it achieves automation and intelligence in behavior identification, risk detection, and response handling during the delivery process. The overall system improves the monitorability, traceability, and emergency response capabilities of the cold chain logistics process. It is suitable for scenarios with extremely high requirements for cold chain continuity, such as food and pharmaceuticals, and has good prospects for industrial promotion and practical application value. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0037] Figure 1 This is a system module diagram of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] For examples, please refer to Figure 1 As shown in the figure, the fresh food cold chain traceability and early warning system based on Internet of Things technology described in this embodiment includes a GPS positioning module, a temperature and humidity sensing module, an electronic fence management module, a parking identification module, a cargo compartment status monitoring module, a data fusion analysis module, a route execution determination module, and an anomaly early warning module.

[0040] The GPS positioning module is used to collect real-time location information of cold chain vehicles;

[0041] Temperature and humidity sensing module, used to collect temperature and humidity data of the refrigerated environment;

[0042] The electronic fence management module is used to set the geofence information for each stop point in the cold chain delivery task;

[0043] The parking recognition module is used to identify actual parking behavior based on changes in vehicle speed and the duration of continuous stillness.

[0044] The cargo compartment status monitoring module is used to detect the open or closed status of the vehicle-mounted cargo compartment;

[0045] The data fusion and analysis module is used to fuse and analyze GPS positioning information, parking recognition results and cargo compartment status data, and output the corresponding credibility score of the stop behavior;

[0046] The route execution determination module, based on the credibility score and the comparison results of the preset route and stop information in the delivery task, determines whether the vehicle has completed the stop task at the designated delivery point.

[0047] If the anomaly warning module fails to complete, it will generate corresponding warning information and send it to the dispatch center.

[0048] The GPS positioning module is used to collect and record the real-time trajectory of cold chain transport vehicles, obtaining data such as the vehicle's latitude and longitude, speed, direction of travel, and timestamp during the delivery process.

[0049] Built-in Global Positioning System (GPS) or BeiDou satellite module, supporting high-frequency sampling (e.g., once every 5 seconds). Supports differential positioning (DGPS) or RTK enhancement, improving positioning accuracy to sub-meter level. Features a signal loss buffering mechanism, automatically storing positioning data in obstructed environments such as tunnels and overpasses, and automatically retransmitting it after signal recovery. Supports API integration with map services (such as Baidu Maps and Gaode Maps) to achieve route visualization and playback. Location information is uploaded to the cold chain cloud platform via cellular networks (e.g., 4G / 5G) or NB-IoT. Provides basic location data for route deviation detection, electronic fence recognition, trajectory backtracking, and transportation compliance review.

[0050] The temperature and humidity sensing module continuously monitors the refrigerated environment inside cold chain transport vehicles, collecting and uploading temperature and humidity data to ensure that the transportation process meets the preservation standards for fresh products. It employs high-precision digital temperature and humidity sensors (such as SHT35 and DHT22), supporting accuracy of ±0.1℃ and ±1.5%RH. The module can be deployed at multiple points based on the temperature distribution characteristics of the refrigerated space to prevent localized temperature measurement distortion. It supports real-time sampling and local caching, with a configurable sampling interval (e.g., 1 minute). It provides an alarm function for abnormal temperatures exceeding limits, immediately pushing a warning when the threshold is exceeded. It can be integrated into the vehicle controller or exist as a standalone Bluetooth / LoRa module. It provides evidence of the integrity of the temperature control chain, offering crucial monitoring support for the "uninterrupted cold chain" requirement and serving as an important data source for traceability and early warning systems.

[0051] The electronic fence management module is used to set virtual geographical boundaries (Geo-fences) for each preset stop (such as distribution stations, stores, and receiving points) in cold chain delivery tasks, enabling intelligent judgment of whether a vehicle "enters" or "leaves" a certain area. It compares GPS positioning data with the set fence latitude and longitude radius. It supports modeling circular fences, rectangular fences, and custom polygonal fences. The fence size is configured according to the actual site conditions (e.g., 50m or 100m radius). Each fence is equipped with logical triggers: fence entry event, fence departure event, and timeout event for not entering. The system supports binding fences to task IDs to prevent path overlap and identification confusion. It can serve as the basic mechanism for determining whether a preset stop has been reached, and can work with other modules to identify whether the delivery has been truly completed.

[0052] The parking recognition module is used to analyze vehicle speed data and inertia changes to determine whether the vehicle has actually "stopped" at a certain location, thereby helping to determine whether a delivery has been made.

[0053] The system determines a vehicle's "stationary" state by analyzing GPS speed values ​​at multiple consecutive time points (e.g., speeds below 5 km / h for more than 30 seconds). Redundancy verification is performed using vehicle inertial measurement unit (IMU) signals or OBD interface information. A "valid parking" event is triggered only if both a "stationary duration threshold" and "location within the fence" conditions are met simultaneously. The module is adaptable to short stops, temporary waiting, and traffic congestion, eliminating false positives through joint time-location analysis. It compensates for the failure of electronic fences to detect parking due to GPS drift or short-term parking, improving the accuracy and robustness of stop detection.

[0054] The cargo compartment status monitoring module is used to monitor the opening / closing behavior of the refrigerated cargo compartment to determine whether loading and unloading actions have occurred, which is an important clue to determine whether the delivery behavior has been actually executed.

[0055] Hall effect sensors, reed switches, or photoelectric switches are installed on the edge of the cargo door to detect changes in the door's opening and closing status. A vibration detection module (accelerometer) can be used to determine if goods are being handled. Event-triggered data uploads are supported, such as "door opening time, closing time, and door opening duration." Only highly sensitive areas (such as inside the store's perimeter fence) can be designated for door opening; opening in other areas is considered abnormal behavior and triggers an alert. Data is uploaded synchronously with GPS location information, achieving location-behavior binding. This can be directly used to verify whether actual loading and unloading operations are taking place at a specific station, serving as an important auxiliary signal to improve the reliability of delivery behavior identification.

[0056] The data fusion and analysis module is used to fuse and analyze GPS positioning information, parking recognition results, and cargo compartment status data, and output a corresponding credibility score for the stopping behavior, specifically:

[0057] Input data types: GPS trajectory data: current location, speed, historical trajectory points, fence entry / exit time; parking behavior data: duration of vehicle stationary state within a fence, speed curve, inertia change; cargo compartment status data: door opening / closing time, opening duration, whether door opening behavior occurred.

[0058] Align the three types of data according to timestamps to ensure that behavioral characteristics within the same time period can be cross-judged.

[0059] Multiple behavioral feature indicators were extracted and standardized, such as GPS dwell time deviation, abnormal door opening frequency, and offset distance between the parking point and the center of the fence.

[0060] The method for extracting the GPS dwell time deviation is as follows: Input GPS trajectory point sequence P = {(lat1,lon1,t1),(lat2,lon2,t2),…, ... n ,lon n ,t n )};lat i Let lon be the latitude of the i-th GPS point. i Let t be the longitude of the i-th GPS point. i The timestamp of the i-th GPS point is used; points with low accuracy (high HDOP, abnormal speed) are filtered out, and the time interval is standardized (e.g., one point every 10 seconds).

[0061] Map GPS track points to actual roads or geographic areas to identify whether vehicles enter / leave a preset geofence. Use map matching algorithms (such as HMM / Viterbi) to correct GPS drift and jump points. Output a set of track points with matching results, marking whether they are within the target geofence: P′={(lat i ,lon i,t i ,inzone)}; inzone indicates that the target is within the fence.

[0062] Iterate through the matching trajectory points to find data segments where `in zone = true`. Determine if the time span of this segment exceeds the minimum dwell threshold (e.g., 30 seconds).

[0063] Assuming the detected values ​​are: actual entry time: Tenter, actual exit time: Texit, calculate the actual dwell time Tactual: Tactual = Texit - Tenter; multiple dwell periods can be accumulated (e.g., repeated entry and exit); obtain the expected dwell time Texpected, and calculate the GPS dwell time deviation value ΔT, expressed as: ΔT = Tactual - Texpected; if: ΔT > 0: dwell time exceeded; ΔT < 0: dwell time insufficient; ΔT = 0: consistent with expectation.

[0064] Door opening frequency anomalies refer to the degree of deviation between the actual number of times the vehicle's cargo compartment door is opened and the system's expected number of openings within a certain cold chain delivery task cycle. They are used to detect abnormal behaviors such as excessive door openings, no doors being opened, or doors being opened in abnormal areas.

[0065] The method for extracting outliers in door opening frequency is as follows: Each time the cargo compartment door is opened, the sensor records one event: D = {(t1, loc1), (t2, loc2), ..., (t m ,loc m Each data entry contains a door opening timestamp t. i Door opening position coordinates loc i (or fenced area ID), m is the total number of coordinates; the pre-set tasks in the system usually include: a list of stop points S = {s1, s2, ..., s z}; Whether each station requires opening the door (default is one door opening per station); Set the expected number of openings as: Nexpected = z; that is, if there are z delivery points, the expected number of door openings is z times, and count the actual number of door openings: Nactual = len(D); Count the total number of actual door opening events. The abnormal value of door opening frequency ΔN is defined as: ΔN = Nactual - Nexpected; If ΔN > 0: suspected of frequent door openings, with goods being taken out illegally or frequent operations; If ΔN < 0: suspected of not completing delivery at the station or not delivering goods; If ΔN = 0: consistent with the task plan, considered normal.

[0066] Offset distance refers to the geographical distance between the vehicle's "actual parking location" at a delivery point and the "preset electronic fence center point" of that site, usually expressed in meters.

[0067] In the task setting of the system, a fence is usually set for each delivery point, and its center point coordinates are: C = (lat c , lon c ); that is, the center of the electronic fence of the site is preset as the most ideal parking position.

[0068] Obtain the parking point coordinates of the vehicle from the parking recognition module: P = (lat p , lon p ); The midpoint, starting point or weighted average position of the parking section can be taken as the representative point.

[0069] Adopt the earth plane approximation method for calculation: Where: Δlat = lat p - lat c ; Δlon = lon p - lon c ; D is the offset distance, with the unit of meter (m); 111,000 means that each degree on the earth is approximately equal to 111 kilometers (rough conversion). The offset threshold can be set: D ≤ 30m: normal approaching point; 30 < D ≤ 100m: edge docking, manual verification is required; D > 100m: suspected non-docking or false residence.

[0070] Normalize the GPS residence time deviation value, the abnormal value of the door opening frequency, and the offset distance between the parking point and the fence center, so that they are all within [0, 1]. According to the normalized GPS residence time deviation value, the abnormal value of the door opening frequency, and the offset distance between the parking point and the fence center, calculate the credibility score of the stop behavior through weighted average summation.

[0071] Compare the obtained credibility score of the stop behavior with the gradient standard threshold. The gradient standard threshold includes the first standard threshold and the second standard threshold, and the first standard threshold is less than the second standard threshold. Compare the credibility score of the stop behavior with the first standard threshold and the second standard threshold respectively;

[0072] If the credibility score of the stop behavior is greater than the second standard threshold, it is determined as a highly credible delivery behavior, indicating that the vehicle's docking behavior at this site is stable, compliant, and reliable, and can be recorded as "effectively completed the task" without intervention, and the score is recommended to be in the green state;

[0073] If the credibility score of the stop behavior is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it is determined as a medium-credible delivery behavior, indicating that the behavior partially meets the expectations but there are slight abnormalities. The system can mark it as "suspicious completion" and recommend dispatching for manual review, which can be used to generate subsequent quality inspection or spot-check tasks;

[0074] If the credibility score of the stopover behavior is less than the first standard threshold, it is judged as a low-credibility delivery behavior, which is suspected of not actually completing the delivery, making a false stop or evading the task. The system should immediately issue an alert and record the abnormal behavior, trigger the traceability intervention process, such as freezing the traceability chain of the order.

[0075] The system compares the preset routes and stop information in the delivery task, and combines the comparison results with a reliability score for comprehensive analysis to determine whether the vehicle has completed the stop task at the designated delivery point. Specifically:

[0076] Preset path data R = {r1, r2, ..., r p}: p represents the total number of path data in the sequence, including the latitude and longitude of each delivery point, the fence radius, and the required dwell time;

[0077] Actual trajectory data P = {(lat1,lon1,t1),(lat2,lon2,t2),…, ...(lat1,lon1,t1),(lat2,lon2,t2),(lat1,lon1,t1),(lat2,lon2,t2),(lat1,lon1,t2),(lat2,lon2,t2),(lat1,lon1,t2),(lat2, n ,lon n ,t n )}: GPS sampling trajectory sequence;

[0078] The credibility score for each stop is Gi∈[0,1];

[0079] Perform map matching on the trajectory sequence and correct GPS jump points;

[0080] Iterate through each stop r I Perform the following judgment:

[0081] Does the trajectory P contain any trajectory points that fall within the fence Ei (a defined radius area)?

[0082] Does the continuous dwell time exceed the set threshold (e.g., >30 seconds)?

[0083] Calculate the trajectory similarity between the actual trajectory and the desired path segment (using DTW or Frechet distance): D dtw The actual DTW distance between the actual and preset trajectories; Sim i ∈[0,1]: Trajectory matching similarity;

[0084] Sim, the trajectory comparison results i A comprehensive analysis is performed with the behavioral credibility score Ci, and the overall score is calculated to define the final path execution credibility F. i :F i =α·Ci+(1-α)·Sim i α∈[0.5,0.9]: The weighting coefficient between credibility score and trajectory similarity (with an emphasis on behavior score);

[0085] The final route execution reliability is compared with a predetermined threshold. If the route execution reliability is greater than or equal to the predetermined threshold, the route is considered successfully executed. This indicates that the vehicle's behavior at all (or major) stops throughout the entire delivery task conforms to the system's set reliability rules. It shows that the vehicle basically completed all stop tasks according to the preset route and requirements, and the delivery behavior is genuine and reliable. The system marks the task path as "validally completed," and it can proceed to the order reconciliation, settlement, and traceability chain generation process. No scheduling intervention or anomaly review is required.

[0086] If the reliability of route execution is less than the predetermined threshold, it is judged as route execution abnormality, indicating that the vehicle has obvious abnormal behavior in this delivery task, such as: missing some stops; not reaching a reasonable dwell time; abnormal frequency of door opening behavior; actual trajectory seriously deviating from the preset route; the system marks the task route as "delivery abnormal or interrupted", which may trigger the following operations: automatically sending an abnormality warning to the dispatch platform; transferring the task to the manual review process; freezing the traceability record or settlement of this batch of orders; and including it in the behavior scoring system for driver or carrier assessment.

[0087] The anomaly warning module automatically generates an anomaly warning when the route execution determination module determines that the delivery vehicle has not completed the specified route or some key stop tasks, and sends the information to the dispatch center for further processing. Anomalies are considered to exist in route execution when the system determines that the route execution reliability is below a set threshold, the reliability score of any important stop is significantly low, or the offset distance between the actual parking point and the preset fence center point exceeds a reasonable range.

[0088] When any of the above-mentioned abnormal situations occur, the anomaly warning module will immediately generate an early warning message containing key information such as the anomaly type, time of occurrence, location, and path score. This early warning information is transmitted to the dispatch center system through the platform's communication interface, or sent to relevant management personnel in real time via a configured message push mechanism (such as API calls, message queues, or mobile push notifications) so that timely measures such as dispatch intervention, task review, vehicle reassignment, or anomaly reporting can be taken.

[0089] Furthermore, each abnormal warning event will be recorded in the system's log database, serving as the basis for subsequent cold chain traceability, task auditing, and performance evaluation. Through this module, the system achieves automatic monitoring, intelligent identification, and real-time feedback at the delivery behavior level, ensuring the standardization and controllability of the cold chain transportation process and significantly improving the security and response efficiency of the overall traceability system.

[0090] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0091] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0092] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A fresh food cold chain traceability and early warning system based on Internet of Things (IoT) technology, characterized in that: It includes a GPS positioning module, a temperature and humidity sensing module, an electronic fence management module, a parking recognition module, a cargo compartment status monitoring module, a data fusion and analysis module, a route execution determination module, and an anomaly warning module; The GPS positioning module is used to collect real-time location information of cold chain vehicles; Temperature and humidity sensing module, used to collect temperature and humidity data of the refrigerated environment; The electronic fence management module is used to set the geofence information for each stop point in the cold chain delivery task; The electronic fence management module supports custom polygon settings for fence areas and configures task numbers, time windows, and docking behavior requirements for each fence. The parking recognition module is used to identify actual parking behavior based on changes in vehicle speed and the duration of continuous stillness. The cargo compartment status monitoring module is used to detect the open or closed status of the vehicle cargo compartment. The cargo compartment status monitoring module collects the open / closed status of the cargo compartment based on the door magnetic sensor, and allows the door to be opened when the vehicle is within the designated electronic fence. If the door is opened outside the preset area, an abnormal warning will be automatically triggered. The data fusion and analysis module is used to fuse and analyze GPS positioning information, parking recognition results and cargo compartment status data, and output the corresponding credibility score of the stop behavior; The data fusion and analysis module specifically includes: acquiring the vehicle's GPS trajectory information, parking recognition data, and cargo compartment status data; aligning the data according to timestamps; extracting stop behavior feature values, including GPS dwell time deviation values, abnormal door opening frequency values, and offset distance between the parking point and the center of the fence; normalizing the feature values ​​and calculating the credibility score of the stop behavior according to the weighted average formula, with the score range from 0 to 1. The route execution determination module, based on a reliability score and a comparison of the preset route and stop information in the delivery task, determines whether the vehicle has completed the stop task at the designated delivery point. Specifically: Preset path data Where p is the total number of path data in the sequence, and the preset path data... Includes the latitude and longitude of each delivery point, fence radius, and required dwell time; GPS track point sequence. ; The credibility score for each stop is Ci∈[0,1]; Map matching is performed on the trajectory sequence to correct GPS jump points; Calculate the trajectory similarity between the actual trajectory and the desired path segment: ; The actual DTW distance between the actual trajectory and the preset trajectory; ∈[0,1]: Trajectory matching similarity; Track comparison results A comprehensive analysis is performed with the behavioral credibility score Ci to define the final path execution credibility. : ; α∈[0.5,0.9]; If the anomaly warning module fails to complete, it will generate corresponding warning information and send it to the dispatch center. The method for extracting GPS dwell time deviation is as follows: Input GPS trajectory point sequence Map GPS track points to actual roads or geographic areas, use map matching algorithms to correct GPS drift and jump points, and output a set of track points with matching results, marking whether they are within the target fence: ; Let i be the latitude of the i-th GPS point. Let be the longitude of the i-th GPS point. Let i be the timestamp of the i-th GPS point. This indicates the location is within the target fence; iterate through the matching trajectory points, find consecutive data segments where `in zone = true`, obtain the actual entry time (Tenter) and actual exit time (Texit), and calculate the actual dwell time. : Obtain the expected dwell time Texpected for the mission, and calculate the GPS dwell time deviation value. The expression is: ; The method for extracting outliers in door opening frequency is as follows: each time the cargo compartment door is opened, the sensor records one event: Each data entry includes a door opening timestamp. Door opening position coordinates m represents the total number of coordinates; the preset tasks include: a list of stop points. ; Does each station require the door to be opened?; Set the expected number of openings as: Nexpected = z; that is, if there are z delivery points, the expected number of openings is z times. Count the actual number of openings: Nactual = len(D1); Count the total number of actual door opening events, and identify outliers in the opening frequency. Defined as: ; The offset distance is obtained as follows: In the task settings, set a fence for each delivery point, and the coordinates of its center point are: The center of the electronic fence at the station is preset as the ideal parking location; the coordinates of the vehicle's parking point are obtained from the parking recognition module. The calculation is performed using the Earth-plane approximation method. ;in: ; ; D is the offset distance; The credibility score of the stopover behavior is compared with the gradient standard threshold, which includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. The credibility score of the stopover behavior is compared with the first standard threshold and the second standard threshold respectively. If the credibility score of the stopover behavior is greater than the second standard threshold, it is judged as a high-credibility delivery behavior; if the credibility score of the stopover behavior is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it is judged as a medium-credibility delivery behavior; if the credibility score of the stopover behavior is less than the first standard threshold, it is judged as a low-credibility delivery behavior. The final path execution reliability is compared with a predetermined threshold. If the path execution reliability is greater than or equal to the predetermined threshold, the path execution is considered successful, indicating that the vehicle's behavior at all stops during the entire delivery task complies with the system's set reliability rules. If the reliability of the route execution is less than the predetermined threshold, it is determined to be an abnormal route execution, indicating that the vehicle has obvious abnormal behavior in this delivery task, and an abnormality warning is automatically sent to the dispatch platform.

Images