Cold chain transportation whole-process quality monitoring method and system based on multi-modal perception
By using multimodal sensing technology and combining temperature field, vibration, gas, and visual data, a comprehensive quality assessment model is constructed, which solves the problem of incomplete monitoring in cold chain transportation and enables real-time early warning and dynamic assessment of changes in cargo quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- QINGDAO HENGXING UNIV OF SCI & TECH
- Filing Date
- 2026-02-27
- Publication Date
- 2026-06-05
AI Technical Summary
Existing cold chain transportation monitoring technologies mainly rely on monitoring a single environmental parameter (temperature), which cannot fully reflect the microenvironment of goods and lacks comprehensive perception of non-temperature factors, resulting in delayed early warnings and an inability to dynamically assess the quality of goods.
By employing a multimodal sensing method, a comprehensive quality assessment model is constructed through the collection and fusion analysis of multi-dimensional data such as temperature field, vibration, gas, and vision, enabling real-time monitoring and multi-level early warning.
It enables real-time monitoring and early warning of cargo quality during cold chain transportation, solving the problems of traditional monitoring systems being unable to fully reflect the environment and having delayed early warnings, and dynamically assessing changes in cargo quality.
Smart Images
Figure CN122155502A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of cold chain transportation monitoring technology, and in particular to a method and system for monitoring the quality of the entire cold chain transportation process based on multimodal perception. Background Technology
[0002] Currently, quality monitoring during cold chain transportation mainly relies on monitoring single environmental parameters (especially temperature). Existing solutions typically involve installing several independent temperature recorders inside the transport vehicle (such as refrigerated trucks or containers) or placing disposable temperature tags inside the cargo packaging to record historical temperature data throughout the transportation process. This data is usually read after transportation to trace and verify whether any over-temperature events occurred during transport. Some advanced systems employ remote temperature monitoring technology based on the Internet of Things (IoT), enabling near real-time temperature uploading and alarm functions.
[0003] These existing technological solutions primarily focus on the "temperature" dimension, resulting in inadequate monitoring and early warning capabilities. First, the interior of cold chain cargo compartments possesses a complex spatial temperature field, making it difficult to comprehensively reflect the actual temperature conditions of the microenvironment within the goods using only a limited number of monitoring points. In particular, they are unable to effectively monitor localized hotspots or chilling injury caused by poor cold air circulation or improper cargo stacking. Second, non-temperature factors such as mechanical vibrations during transportation, fluctuations in humidity and gas composition due to frequent door opening and closing, and light exposure also significantly impact the quality of specific goods (such as precision instruments, fruits and vegetables, and flowers). Existing systems lack the comprehensive perception capability for these critical factors. Third, post-event or real-time alarms based on single thresholds cannot assess or predict deterioration in cargo quality, failing to provide early warnings before irreversible quality loss occurs.
[0004] Therefore, in response to the problems mentioned above, this invention proposes a method and system for monitoring the quality of the entire cold chain transportation process based on multimodal perception. Summary of the Invention
[0005] To overcome the problems of existing cold chain transportation monitoring technologies, such as limited monitoring dimensions, inability to accurately reflect the comprehensive environment of goods, delayed early warnings, and inability to perform dynamic quality prediction, this invention proposes a method and system for full-process quality monitoring of cold chain transportation based on multimodal perception. This method collects multi-dimensional environmental and cargo status data, including temperature field, vibration, gas, and visual data, and performs fusion analysis and environmental modeling to achieve real-time and comprehensive monitoring, evaluation, and early warning of cargo quality throughout the transportation process.
[0006] The technical solution of this invention is: a method for monitoring the quality of the entire cold chain transportation process based on multimodal sensing, comprising the following steps: S1, through sensing units deployed on the cold chain transport carrier, collects real-time sensing data of the transport environment. The sensing data includes at least the following: temperature data (preferably temperature measurement range -25℃ to +50℃, accuracy ±0.3℃) collected by a temperature sensor array consisting of at least 9 temperature probes deployed at the top, middle, bottom, and four corners of the cargo compartment; humidity data (preferably measurement range 0%RH-100%RH, accuracy ±2%RH) collected by a humidity sensor; vibration and attitude data collected by a triaxial accelerometer (preferably range ±16g) and a gyroscope (preferably range ±2000° / s); gas concentration data collected by a gas concentration sensor capable of detecting ethylene, carbon dioxide, and ammonia; and cargo surface image data collected by an image acquisition module (preferably capture interval of 15-60 minutes, triggering immediate snapshot when the compartment door is detected to be open or abnormal vibration is detected). S2, after preprocessing and synchronizing the sensed data with timestamps, transmits it to the monitoring platform in real time or near real time through the wireless communication unit; S3, within the monitoring platform, performs fusion analysis on the received sensing data, specifically including: S31 performs spatial interpolation calculations on multi-point temperature data to generate a three-dimensional temperature field distribution model inside the transport compartment, and calculates the dew point temperature distribution inside the compartment by combining humidity data. S32 performs time-frequency analysis on vibration and attitude data, extracts characteristic parameters that characterize transportation stability, and identifies sudden braking, frequent turning, or abnormal bump events by combining historical route GIS information. S33, using computer vision models to analyze the quality deterioration characteristics of cargo surface image data, and identify visual anomalies including ice crystal condensation, color changes, shrinkage or signs of decay. S34 compares specific gas concentration data with preset dynamic thresholds based on different goods categories (such as fresh fruits and vegetables, frozen meat, and medicines) to monitor abnormal concentrations of ethylene, carbon dioxide, or ammonia. S4. Based on the fusion analysis results of step S3, a comprehensive quality assessment model for the transported batch of goods is constructed. This model weights and fuses the three-dimensional temperature field uniformity index, cold accumulated temperature index, vibration cumulative damage index, visual deterioration index and gas anomaly index to output the real-time quality score and predicted remaining shelf life of the current batch. S5. When the quality score output by the comprehensive quality assessment model is lower than the dynamic threshold set according to the value and shelf life of the goods (e.g., a score below 70 triggers an early warning), or when any single indicator exceeds the safety threshold, a multi-level risk warning message containing Level 1 (serious), Level 2 (warning), and Level 3 (prompt) is generated. The warning message includes at least the warning level, the type of abnormal indicator, the location and time of occurrence, and handling suggestions (e.g., "adjust refrigeration settings", "check stacking", "prioritize delivery"), and is pushed to the relevant management and carrier terminals. S6 generates a transportation quality report based on the perception data and analysis results throughout the entire process. The report includes complete environmental parameter curves from the loading point to the unloading point, key event records, quality score change trends, and details of abnormal events.
[0007] Preferably, in step S1, the sensing unit further includes a light intensity sensor (range 0-2000 lux) for collecting light intensity data during transportation or loading and unloading, especially for photosensitive pharmaceuticals or some fruits and vegetables. This data will be used in conjunction with image data to assess the risk of fading of goods or degradation of chemical components due to improper lighting.
[0008] Preferably, in step S3, the construction of the three-dimensional temperature field distribution model is as follows: based on the physical location coordinates of the temperature sensor, the Kriging interpolation algorithm is used to generate a continuous temperature distribution surface inside the cargo compartment, and combined with the three-dimensional model of the cargo compartment for visualization rendering. The temperature rendering color gradation can intuitively distinguish areas that are higher or lower than the set target temperature.
[0009] Preferably, in step S3, identifying sudden braking, frequent steering, or abnormal bump events specifically includes: calculating the vector composite amplitude of accelerometer data; marking it as a suspected severe vibration event when it exceeds 0.8g within three consecutive sampling periods; judging the rate of change of the carrier's attitude angles (pitch angle, roll angle) in combination with gyroscope data; confirming it as a sudden steering or roll event when it changes by more than 15° within 1 second; and recording the timestamp, duration, and maximum intensity of the event in association with the corresponding GIS coordinates, and highlighting it on the map trajectory.
[0010] Preferably, in step S4, the cold accumulated temperature index is calculated as follows: taking the reference safe temperature T0 of a specific type of cargo as the lower limit (e.g., -18℃ for frozen tuna, 2℃ for strawberries), the difference between the average temperature of the area where the cargo is located in the real-time three-dimensional temperature field model and the reference safe temperature is integrated over time. The integration interval is from the loading time to the current time. When the average temperature is lower than T0, the difference is zero and is not included in the negative accumulation.
[0011] This invention proposes a multimodal sensing-based end-to-end quality monitoring system for cold chain transportation, comprising an on-board monitoring terminal, a remote cloud monitoring platform, and a user interaction terminal. The vehicle-mounted monitoring terminal includes: The sensing module integrates a temperature sensor array, a humidity sensor, a three-axis accelerometer and gyroscope, a gas sensor, and an image acquisition module, which is used to collect sensing data of the transportation environment. The image acquisition module is a wide-angle camera with an infrared fill light function and a field of view of not less than 120°. The edge computing and communication module is used to preprocess and timestamp the collected data, and upload the data to the remote cloud monitoring platform via a wireless network; The positioning module is used to acquire the real-time location information of the transport vehicle and upload it synchronously with the sensing data; The power management module is used to supply power to the various modules of the terminal. The remote cloud monitoring platform includes: The data receiving and storage module is used to receive and store multimodal perception data uploaded from multiple vehicle terminals; The data fusion and analysis module includes a temperature field analysis unit, a vibration event analysis unit, a visual analysis unit, and a gas analysis unit. Each unit processes the corresponding type of sensing data in parallel, and the data fusion and analysis module performs spatiotemporal alignment and correlation analysis. The quality assessment module is used to build a comprehensive quality assessment model for transported batches of goods, and outputs the real-time quality score and predicted remaining shelf life of the current batch. The early warning module is used to generate multi-level risk warning information when the quality score output by the comprehensive quality assessment model is lower than the threshold, or when any single indicator exceeds the safety threshold. It includes an adjustable rule base, which allows users to preset or adjust the warning thresholds and warning level strategies of various indicators according to different cargo categories, transportation seasons and routes. The report generation service generates transportation quality reports based on the perception data and analysis results throughout the entire process. The user interaction terminal is a web or mobile application used to display real-time transportation status, early warning information, quality assessment results, and traceability reports to administrators, carriers, or cargo owners.
[0012] The beneficial effects of this invention are: 1. This invention constructs a three-dimensional temperature field model by spatial interpolating multi-point temperature data, and combines the spatial location of vibration events with the analysis of images and gas data. It can integrate and analyze the spatiotemporal correlation between different modal data, thereby generating an environmental profile that reflects the comprehensive condition of the microenvironment in which the cargo is actually located, solving the problem that traditional point monitoring cannot truly reflect the spatial environment inside the cargo compartment.
[0013] 2. This invention sets dynamic threshold rules based on multi-dimensional data and utilizes edge computing and wireless communication technologies to achieve real-time data uploading and instant event judgment. This enables the system to identify and issue multi-level warnings in real time for risks such as local overheating, abnormal vibration, accumulation of harmful gases, and visually visible deterioration during transportation. This transforms traditional post-event traceability into in-event intervention, thereby solving the problem of delayed warnings.
[0014] 3. This invention constructs a comprehensive quality assessment model that integrates multiple indicators such as temperature uniformity, cold accumulated temperature, vibration damage, visual deterioration, and gas anomalies. By introducing time factors and initial state, it achieves dynamic quantitative scoring of goods quality and scientific prediction of remaining shelf life. This shifts the focus of monitoring from whether the environment exceeds standards to how quality changes, solving the problem that existing technologies cannot dynamically assess quality. Attached Figure Description
[0015] Figure 1 The diagram shown is a schematic representation of the system framework of the present invention. Figure 2 The diagram shown illustrates the strawberry cold chain transportation process of this invention. Figure 3 The diagram shown illustrates the frozen tuna transportation process of this invention. Detailed Implementation
[0016] 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 some embodiments of the present invention, but 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.
[0017] Please see Figure 1 This invention provides an embodiment: a cold chain transportation end-to-end quality monitoring system based on multimodal perception. In this embodiment, the vehicle-mounted monitoring terminal will be described in detail: This module is not simply a stack of sensors, but an integrated design specifically for the cold chain transportation environment. Its temperature sensor array employs at least nine probes, arranged in three layers (top, middle, and bottom) and strategically positioned along the front, middle, rear, and sides of the cargo compartment to create a gridded monitoring capability of the internal temperature field. Each probe uses a high-precision digital temperature sensor with a measurement range of -25℃ to +50℃ and an accuracy of ±0.3℃, ensuring accurate measurements in various temperature zones, including frozen and refrigerated conditions. Humidity and gas sensors (detecting ethylene, CO2, and NH3) are installed in the middle of the cargo compartment to monitor the overall environmental level. A three-axis accelerometer and gyroscope module is mounted on the cargo compartment floor or frame to sense vibrations and attitude changes during transport. The image acquisition module uses an industrial camera with a wide-angle lens and supplemental lighting, installed in the upper front corner of the cargo compartment, covering the main cargo stacking area. All sensor data is collected by the edge computing and communication module, which is responsible for preprocessing the raw data such as filtering, calibration, and timestamp synchronization, and then uploading the packaged data to the cloud in real time via 4G / 5G or NB-IoT networks according to a preset period (such as every 5 minutes) or event trigger (such as opening a door or severe vibration).
[0018] In this embodiment, the remote cloud monitoring platform will be described in detail: After acquiring data from numerous vehicle terminals, the data receiving and storage module stores it in a time-series database (for storing timestamped sensor readings) and a relational database (for storing event and configuration data). Then, the data fusion and analysis module is activated. This module contains multiple parallel analysis units: the temperature field analysis unit uses the Kriging spatial interpolation algorithm to generate a continuous temperature distribution cloud map of the entire cargo compartment's interior space in real time, based on uploaded discrete-point temperature data and combined with the physical three-dimensional model of the cargo compartment. It also calculates temperature uniformity indicators (such as the standard deviation of spatial temperature) and identifies "cold spots" (where excessively low temperatures may cause frost damage) and "hot spots" (where excessively high temperatures may cause spoilage); the vibration event analysis unit... The system performs time-frequency domain analysis on acceleration and gyroscope data, sets intelligent thresholds (e.g., horizontal acceleration exceeding 0.5g for 200ms is considered emergency braking, and attitude angle change rate exceeding 15° / second is considered sharp turning), automatically identifies, classifies, and records abnormal mechanical events during transportation, and associates them with GIS map locations. The visual analysis unit calls a pre-trained deep learning model to automatically analyze uploaded cargo images, identifying visual features of quality deterioration such as abnormal color, mold, wilting, or frost, and providing confidence scores. The gas analysis unit continuously monitors the concentration trend of specific gases, triggering an alert when the concentration exceeds a threshold set according to the cargo type (e.g., ethylene concentration exceeding 1ppm in strawberry transportation). Finally, the comprehensive quality assessment model receives the index results from each analysis unit, combines the transportation time with specific biochemical and kinetic parameters of the cargo type (e.g., respiration rate, enzyme activity change model), and uses a weighted fusion algorithm to dynamically calculate a comprehensive quality score (Q value, 0-100 points) and predict the remaining shelf life. The early warning module continuously monitors various raw indicators and the overall Q value. Once a rule is triggered, it immediately generates multi-channel early warning information that includes the level, content, and suggestions.
[0019] In this embodiment, the remote cloud monitoring platform will be described in detail: The user interface (web and mobile app) provides users with a monitoring interface. Managers can view the real-time location and status of all vehicles en route on a map (color-coded for healthy, warning, and abnormal), and can also view a detailed dashboard for a specific transport, which dynamically displays a 3D model of the temperature field, data curves from various sensors, quality score trend charts, warning logs, and images of the cargo. After the transport is completed, the system can automatically generate a monitoring report containing all key data and events.
[0020] Please see Figure 2 The present invention provides Embodiment 1: This embodiment describes the transportation of a batch of fresh strawberries from a planting base to a large supermarket 2,000 kilometers away. The entire journey is expected to take 96 hours, using refrigerated trucks with a set temperature of 2°C. Strawberries are extremely sensitive to temperature fluctuations, easily suffering from frost damage (<0°C) and softening and rotting (>5°C). At the same time, the ethylene produced by their respiration accelerates their ripening and spoilage.
[0021] Equipment layout: In this embodiment, nine temperature probes (T1-T9) are arranged inside a 12-meter-long refrigerated cargo compartment according to the above scheme. Among them, T1, T2, and T3 are located on the upper, middle, and lower front layers, close to the door; T4, T5, and T6 are located in the middle; T7, T8, and T9 are located at the rear. The gas concentration sensor is installed on the side of the middle shelf, the camera is installed on the front top, facing the stack of goods, and the accelerometer is installed in the middle of the cargo compartment floor.
[0022] A transportation task was created for this batch on the cloud platform, with "strawberries" selected as the cargo category. The system automatically loaded preset parameters: target temperature 2±0.5℃, humidity 85-95%, ethylene warning threshold 1ppm, and CO2 warning threshold 5000ppm. The comprehensive quality assessment model used strawberry-specific weighting coefficients, with temperature uniformity having the highest weight, and the cold accumulated temperature (TTI) calculation baseline temperature T0 set to 2℃.
[0023] Transportation process: (1) During the smooth driving phase, the system's three-dimensional temperature field model showed that the temperature in most areas remained stable at 1.8-2.2℃. However, during a drive on a mountain road, the model showed that the temperature in the lower rear layer (point T9) slowly rose to 3.5℃, forming a local "hot spot". The system analysis indicated that the reason was that this area was close to the return air vent of the refrigeration unit, and the excessively dense stacking of goods caused poor airflow.
[0024] (2) When the transport reached 36 hours, the gas concentration sensor detected that the ethylene concentration rose to 1.2 ppm. At this time, the warning module triggered a level 2 warning. At the same time, the visual analysis unit analyzed the latest captured image and showed that some strawberry fruits showed signs of slight softening on the surface (confidence level 85%).
[0025] (3) Based on temperature fluctuations (the hot spot lasted for 4 hours), increased ethylene concentration, and initial visual deterioration, the real-time Q value output by the comprehensive quality assessment model decreased from the initial 95 points to 72 points. The model predicts that if transportation continues under the current conditions, the remaining shelf life will be less than 24 hours, which is lower than the planned sales window after arrival.
[0026] (4) At this time, the cloud platform generates a level one warning, suggesting "inspect and clear the stack of goods at the rear, and strengthen ventilation. Prioritize sorting upon arrival." The warning information is pushed to the driver's mobile app and the rear dispatch center in real time. The driver stops in a safe area to check and adjusts the placement of the goods at the rear. After the adjustment, the temperature at the rear gradually drops to 2.5℃.
[0027] While using this invention for transportation, a comparative study was conducted, specifically comparing a traditional method using a single-point temperature recorder (placed in the middle of the cargo compartment) and disposable temperature tags. Throughout the transportation process, the recorder showed the temperature remained around 2°C. Upon arrival at the destination, during unloading, it was discovered that approximately 10% of the strawberries at the rear had noticeably softened and partially rotted, resulting in significant losses. The temperature tags indicated that the temperature was within limits, but the cause of the loss could not be explained.
[0028] Please see Figure 3 The present invention provides embodiment 2: This embodiment describes the transportation of frozen tuna from a fishing port to overseas via sea container. The temperature must be kept stable below -50°C throughout the entire journey. Any temperature fluctuation will affect the cell structure and taste, causing frostbite. At the same time, the ship is constantly rocking during sea transport, but violent rolling or pitching may cause the cargo to shift or even damage the container.
[0029] Equipment layout: Inside the 12-meter refrigerated container, the temperature sensor array has been increased to 12 points, focusing on monitoring the temperature near six sides of the container. The accelerometer and gyroscope modules have been upgraded with higher sensitivity settings to capture complex ship movements. Since it involves deep-sea frozen goods, gas monitoring primarily focuses on whether there is ammonia leakage (a sign of refrigerant anomalies). The imaging camera is equipped with stronger infrared illumination to adapt to the low-temperature, dark environment.
[0030] The target temperature is set to -50℃. Vibration monitoring is set to two levels: continuous routine vibration is analyzed for fatigue accumulation; instantaneous impact events (such as cargo drops) are analyzed with an acceleration threshold >5g. The quality model weights are tilted towards temperature stability and vibration accumulation energy.
[0031] Transportation process: (1) When crossing the equatorial region, the ambient temperature is very high. The three-dimensional temperature field model shows that although the average temperature is maintained at -50.5℃, the temperature gradient is large near the inner wall on the side close to the door (points T2 and T5), and the temperature in the outer area has risen to -48℃.
[0032] (2) When the ship encounters wind and waves, it experiences periodic rolling that lasts for several hours. The vibration analysis unit not only records the event, but also calculates the cumulative vibration energy acting on the cargo during this period (based on the acceleration power spectral density integral) through an algorithm and inputs it into the quality model.
[0033] (3) At a certain moment, the system simultaneously detected an abnormal temperature rise (to -45°C) at a specific point inside the chamber and a slight increase in the ammonia sensor reading. The comprehensive diagnostic module judged it as "high probability of slight local refrigerant leakage" rather than simply external heat intrusion.
[0034] (4) In response to the abnormal increase in local temperature, the system issued a maintenance warning, reminding the container to pay attention to the insulation performance of its door seal in subsequent use. The cumulative vibration data became key objective evidence in cargo damage insurance claims.
[0035] While using this invention for transportation, a comparative method was also employed. Specifically, a simple temperature recording device built into the container was used, typically with only 1-2 temperature measurement points, which could only record whether the average temperature met the standard. Vibration was completely unmonitored. If, upon arrival at port and opening the container, localized darkening of the tuna's color (signs of frostbite) or damage to the container was found, disputes could easily arise between the shipping company and the cargo owner regarding the cause (whether it was a temperature issue or a handling issue), and evidence would be lacking.
[0036] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.
Claims
1. A method for end-to-end quality monitoring of cold chain transportation based on multimodal perception, characterized in that, It includes the following steps: S1, through sensing units deployed on the cold chain transport carrier, collects sensing data of the transport environment in real time, including at least temperature data collected by temperature sensors, humidity data collected by humidity sensors, vibration and attitude data collected by triaxial accelerometers and gyroscopes, specific gas concentration data collected by gas sensors, and cargo surface image data collected by image acquisition modules. S2, after preprocessing and synchronizing the sensed data with timestamps, transmits it to the remote cloud monitoring platform in real time or near real time through the wireless communication unit; S3, within the remote cloud monitoring platform, performs fusion analysis on the received sensing data, specifically including: S31 performs spatial interpolation calculations on multi-point temperature data to generate a three-dimensional temperature field distribution model inside the transport compartment, and calculates the dew point temperature distribution inside the compartment by combining humidity data. S32 performs time-frequency analysis on vibration and attitude data, extracts characteristic parameters that characterize transportation stability, and identifies sudden braking, frequent turning, or abnormal bump events by combining historical route GIS information. S33, using computer vision models to analyze the quality deterioration characteristics of cargo surface image data, and identify visual anomalies including ice crystal condensation, color changes, shrinkage or signs of decay. S34 compares specific gas concentration data with preset thresholds to monitor abnormal concentrations of ethylene, carbon dioxide, or ammonia. S4. Based on the fusion analysis results of step S3, a comprehensive quality assessment model for the transported batch of goods is constructed. This model weights and fuses the three-dimensional temperature field uniformity index, cold accumulated temperature index, vibration cumulative damage index, visual deterioration index and gas anomaly index to output the real-time quality score and predicted remaining shelf life of the current batch. S5. When the quality score output by the comprehensive quality assessment model is lower than the threshold, or any single indicator exceeds the safety threshold, a multi-level risk warning message is generated. The warning message includes at least the warning level, the type of abnormal indicator, the location and time of occurrence, and handling suggestions, and is pushed to the relevant management and carrier terminals. S6 generates a transportation quality report based on the perception data and analysis results throughout the entire process. The report includes complete environmental parameter curves from the loading point to the unloading point, key event records, quality score change trends, and details of abnormal events.
2. The method for end-to-end quality monitoring of cold chain transportation based on multimodal perception as described in claim 1, characterized in that: In step S1, the sensing unit also includes an illuminance sensor, which is used to collect light intensity data during transportation or loading and unloading, and to work with image data to assess the quality impact on photosensitive goods.
3. The method for end-to-end quality monitoring of cold chain transportation based on multimodal perception as described in claim 1, characterized in that: In step S3, the construction of the three-dimensional temperature field distribution model is specifically as follows: based on the physical location coordinates of the temperature sensor, the Kriging interpolation algorithm is used to generate a continuous temperature distribution surface inside the cargo compartment, and then combined with the three-dimensional model of the cargo compartment for visualization rendering.
4. The method for end-to-end quality monitoring of cold chain transportation based on multimodal perception as described in claim 1, characterized in that: In step S3, identifying sudden braking, frequent steering, or abnormal bump events specifically includes: calculating the vector composite amplitude of accelerometer data, marking it as a suspected event when it exceeds a first threshold, judging the attitude angle change rate by combining gyroscope data, confirming it as a sudden steering or roll event when it exceeds a second threshold, and recording the timestamp of the event with GIS coordinates.
5. The method for end-to-end quality monitoring of cold chain transportation based on multimodal perception according to claim 1, characterized in that: In step S4, the cold accumulated temperature index is calculated as follows: taking the reference safe temperature of a specific type of cargo as the lower limit, the difference between the average temperature of the area where the cargo is located in the real-time three-dimensional temperature field model and the reference safe temperature is integrated over time.
6. A multimodal sensing-based end-to-end cold chain transportation quality monitoring system, used to implement the multimodal sensing-based end-to-end cold chain transportation quality monitoring method according to any one of claims 1-5, characterized in that, Including: The sensing module integrates a temperature sensor array, a humidity sensor, a three-axis accelerometer and gyroscope, a gas sensor, and an image acquisition module, which is used to collect sensing data of the transportation environment. The edge computing and communication module is used to preprocess and timestamp the collected data, and upload the data to the remote cloud monitoring platform via wireless network; The power management module is used to supply power to the various modules of the terminal. The data receiving and storage module is used to receive and store multimodal perception data uploaded from multiple vehicle terminals; The data fusion and analysis module is used to perform fusion and analysis on the received sensing data; The quality assessment module is used to build a comprehensive quality assessment model for transported batches of goods, and outputs the real-time quality score and predicted remaining shelf life of the current batch. The early warning module is used to generate multi-level risk warning information when the quality score output by the comprehensive quality assessment model is lower than the threshold, or when any single indicator exceeds the safety threshold. The report generation service generates transportation quality reports based on the perception data and analysis results throughout the entire process. The user interaction terminal is a web or mobile application used to display real-time transportation status, early warning information, quality assessment results, and traceability reports to administrators, carriers, or cargo owners.
7. The cold chain transportation end-to-end quality monitoring system based on multimodal perception according to claim 6, characterized in that: The vehicle-mounted multimodal monitoring terminal also includes a positioning module, which is used to acquire the real-time location information of the transport vehicle and upload it synchronously with the sensing data.
8. The cold chain transportation end-to-end quality monitoring system based on multimodal perception according to claim 6, characterized in that, The data fusion analysis module specifically includes: a temperature field analysis unit, a vibration event analysis unit, a visual analysis unit, and a gas analysis unit. Each unit processes the corresponding type of sensing data in parallel, and the data fusion analysis module performs spatiotemporal alignment and correlation analysis.
9. The cold chain transportation end-to-end quality monitoring system based on multimodal perception according to claim 6, characterized in that: The early warning module includes an adjustable rule base, which allows users to preset or adjust the early warning thresholds and early warning levels for various indicators based on different cargo categories, transportation seasons, and routes.
10. The cold chain transportation end-to-end quality monitoring system based on multimodal perception according to claim 6, characterized in that: The image acquisition module is a wide-angle camera with infrared fill light function.