Deep learning-based frozen shrimp transportation intelligent early warning method and system

Abstract

The invention discloses an intelligent early warning method and system for frozen shrimp transportation based on deep learning, and belongs to the technical field of cold-chain logistics and intelligent monitoring. According to the method, a multi-dimensional data analysis model is constructed by using a deep learning technology, and intelligent monitoring and anomaly prediction of a transportation environment are realized. The system is composed of a data acquisition module, a deep learning analysis module and an early warning decision module, can automatically identify potential risks (such as temperature abnormity and equipment faults) in the transportation process, and timely notifies related personnel through a multi-level early warning mechanism (such as short messages, mails and system popup windows). Through self-learning and optimization of the deep learning model, the accuracy and the real-time performance of early warning are improved, the loss risk of frozen shrimps in the transportation process is effectively reduced, the product quality is guaranteed, meanwhile, the intelligent level of cold-chain logistics is improved, and remarkable economic benefits and application values are achieved.

Description

Technical Field

[0001] The present invention relates to a deep learning-based intelligent early warning method and system for frozen shrimp transportation, belonging to the technical field of cold chain logistics and intelligent monitoring. Background Art

[0002] With the continuous improvement of people's living standards and the booming development of fresh food e-commerce, the consumption market of frozen shrimp is showing an increasing trend. The quality assurance of frozen shrimp during transportation has become an important research and practice area.

[0003] Traditional frozen shrimp transportation relies primarily on refrigeration equipment to maintain a low temperature. However, this approach has numerous limitations. For one thing, environmental factors during transportation are difficult to accurately monitor, and simple temperature control alone cannot guarantee optimal frozen shrimp quality. For example, temperature fluctuations can be unavoidable due to differences in refrigeration equipment performance, varying road conditions along the transportation route, and environmental influences. Even small temperature fluctuations can compromise the quality of frozen shrimp, affecting the freshness, taste, and nutritional value of the meat. Furthermore, the impact of other critical factors during transportation, such as humidity, vibration, and transportation time, is often overlooked. High humidity can cause frost to form on the packaging surface, compromising the refrigeration efficiency of the refrigeration equipment and the quality of the shrimp. Excessive vibration during transportation can damage the shrimp's integrity, affecting its appearance and quality. Long transportation periods can cause a series of physical and chemical changes in the shrimp, but traditional transportation methods make it difficult to comprehensively and accurately monitor and analyze these changes.

[0004] Furthermore, traditional frozen shrimp transportation management is largely based on manual experience, lacking the systematic analysis and utilization of large amounts of transportation data. This means that any anomalies during transportation cannot be promptly detected and addressed, potentially resulting in the loss of the entire shipment. Furthermore, the lack of in-depth analysis of historical transportation data makes it difficult to predict potential problems and implement effective preventive measures.

[0005] Furthermore, the current logistics industry lacks a comprehensive intelligent solution for frozen shrimp transportation monitoring and early warning systems, integrating IoT technology and big data analytics. Some existing monitoring systems only provide basic temperature data, with limited functionality. They fail to comprehensively consider multiple influencing factors and fail to establish a complete chain from data collection, transmission, storage, to intelligent analysis and early warning. Summary of the Invention

[0006] In order to solve the problems of incomplete quality monitoring and untimely early warning during the existing frozen shrimp transportation process, the present invention provides an intelligent early warning method and system for frozen shrimp transportation based on deep learning. The technical solution is as follows:

[0007] The present invention provides an intelligent early warning method for frozen shrimp transportation, comprising:

[0008] Step 1: Obtain environmental data during frozen shrimp transportation and perform preprocessing;

[0009] Step 2: Use a deep learning network to perform real-time analysis and prediction of the environmental data. When the analysis and prediction results are greater than a preset threshold, an intelligent early warning mechanism is triggered.

[0010] The analysis and forecasting process includes:

[0011] (1) Constructing a mutual information network, which is represented by:

[0012]

[0013] Among them, P XY (x,y) is the joint distribution; P X (x)P Y (y) is the product of marginal distributions;

[0014] (2) Calculate the lower bound of the mutual information of each modal environment data, expressed as:

[0015]

[0016] Among them, T θ (x,y) represents a neural network, represents the calculation of the expectation of the joint sample, Indicates the expectation of calculating marginal samples;

[0017] (3) Define the multidimensional feature quality vector, expressed as:

[0018] Q=[q1,q2,…q n ]

[0019] Among them, q1,q2,…q n Indicates data quality parameters;

[0020] (4) Calculate quality scores;

[0021] First, the dot product of the multidimensional feature quality vector Q and the weight vector W is calculated:

[0022]

[0023] The weight vector W is obtained by AHP;

[0024] The original score S raw The final quality score is obtained by mapping it to the (0,1) interval through the Sigmoid function:

[0025]

[0026] (5) Heterogeneous data collected by sensors Perform multimodal alignment calculation:

[0027] First, the mutual information of the two modal data is calculated through the mutual information network. The calculation formula is:

[0028]

[0029] Among them, X i and X j are the sensor data of two modes respectively; T(x i ,x j ) represents the statistical function of the mutual information network, represents the joint distribution, represents the marginal distribution product;

[0030] Constructing a cross-modal attention matrix:

[0031]

[0032] Where τ represents the temperature coefficient;

[0033] Compute feature space projection:

[0034]

[0035] Among them, W i represents the private projection matrix of modality i, V j represents the shared value transformation matrix of mode j;

[0036] (6) Dynamically weighted fusion is performed on the feature space projection, and the calculation formula is:

[0037]

[0038] in, Represents the modal alignment feature, all modes are projected into a unified dimension d; quality_score (Q i ) represents the quality score of the i-th modal quality vector, expressed as s i ;

[0039] Calculate dynamic weights and score the quality of each modality s i Perform exponential weighting + normalization:

[0040]

[0041] The weight w iApplied to aligned features Get the final fusion result Y:

[0042]

[0043] (7) Threshold calculation is performed based on the sliding window statistics method. The calculation formula is:

[0044]

[0045] in, represents the mean value of Y in the time window [tw,t], represents the standard deviation of Y in the time window [tw,t], k represents the sensitivity coefficient, and w represents the length of the sliding window;

[0046] (8) The final fusion result Y and the threshold threshold t Make comparisons and trigger different levels of warnings.

[0047] Optionally, the quality data in the multidimensional feature quality vector includes: SNR, sampling rate, historical error rate, data integrity, timeliness, neighborhood consistency, environmental interference resistance, hardware health, signal entropy value, drift coefficient, failure probability, and calibration status.

[0048] Optionally, the environmental data includes: temperature, humidity, position, and acceleration.

[0049] Optionally, the deep learning network includes: a convolutional neural network and a long short-term memory network.

[0050] The present invention provides an intelligent early warning system for frozen shrimp transportation, comprising:

[0051] Sensor modules for collecting environmental data during transportation;

[0052] A data processing module uses a deep learning network to perform real-time analysis and prediction of the environmental data. When the analysis and prediction results are greater than a preset threshold, an intelligent early warning mechanism is triggered;

[0053] The analysis and forecasting process includes:

[0054] (1) Constructing a mutual information network, which is represented by:

[0055]

[0056] Among them, P XY (x,y) is the joint distribution; P X (x)P Y (y) is the product of marginal distributions;

[0057] (2) Calculate the lower bound of the mutual information of each modal environment data, expressed as:

[0058]

[0059] Among them, T θ (x,y) represents a neural network, represents the expectation of calculating the joint sample, Indicates the expectation of calculating marginal samples;

[0060] (3) Define the multidimensional feature quality vector, expressed as:

[0061] Q=[q1,q2,…q n ]

[0062] Among them, q1,q2,…q n Indicates data quality parameters;

[0063] (4) Calculate quality scores;

[0064] First, the dot product of the multidimensional feature quality vector Q and the weight vector W is calculated:

[0065]

[0066] The weight vector W is obtained by AHP;

[0067] The original score S raw The final quality score is obtained by mapping it to the (0,1) interval through the Sigmoid function:

[0068]

[0069] (5) Heterogeneous data collected by sensors Perform multimodal alignment calculation:

[0070] First, the mutual information of the two modal data is calculated through the mutual information network. The calculation formula is:

[0071]

[0072] Among them, X i and X j are the sensor data of two modes respectively; T(x i ,x j ) represents the statistical function of the mutual information network, represents the joint distribution, represents the marginal distribution product;

[0073] Constructing a cross-modal attention matrix:

[0074]

[0075] Where τ represents the temperature coefficient;

[0076] Compute feature space projection:

[0077]

[0078] Among them, W i represents the private projection matrix of modality i, V j represents the shared value transformation matrix of mode j;

[0079] (6) Dynamically weighted fusion is performed on the feature space projection, and the calculation formula is:

[0080]

[0081] in, Represents the modal alignment feature, all modes are projected into a unified dimension d; quality_score (Q i ) represents the quality score of the i-th modal quality vector, expressed as s i ;

[0082] Calculate dynamic weights and score the quality of each modality s i Perform exponential weighting + normalization:

[0083]

[0084] The weight w i Applied to aligned features Get the final fusion result Y:

[0085]

[0086] (7) Threshold calculation is performed based on the sliding window statistics method. The calculation formula is:

[0087]

[0088] in, represents the mean value of Y in the time window [tw,t], represents the standard deviation of Y in the time window [tw,t], k represents the sensitivity coefficient, and w represents the length of the sliding window;

[0089] (8) The final fusion result Y and the threshold threshold t Make comparisons and trigger different levels of warnings.

[0090] Optionally, the sensor module includes: a temperature sensor, a humidity sensor, a position sensor, and an acceleration sensor.

[0091] Optionally, the intelligent early warning mechanism includes:

[0092] Send early warning information to managers via SMS, email, and mobile app push messages;

[0093] Automatically control the environmental control equipment of the transport container and automatically adjust to ensure the transportation environment of frozen shrimp.

[0094] Optionally, the preset thresholds also include: temperature thresholds, humidity thresholds, and acceleration thresholds for vibration and bumping set according to different transportation stages.

[0095] Optionally, the system further includes: a user interface.

[0096] Optionally, the data processing module uses a distributed storage system to store and manage environmental data.

[0097] The beneficial effects of the present invention are:

[0098] This invention utilizes deep learning technology to analyze multi-source sensor data in real time and, by constructing a multimodal data fusion model, accurately identifies abnormal fluctuations in the transportation environment. Compared to traditional threshold alarm methods, this invention combines historical data with real-time status to achieve early risk warnings, significantly reducing the risk of quality deterioration due to temperature control failure or mechanical damage, and ensuring the freshness and safety of frozen shrimp. This invention utilizes a multimodal alignment algorithm based on mutual information estimation to effectively integrate heterogeneous sensor data, eliminate device differences and noise interference, and improve the accuracy and consistency of monitoring data.

[0099] Through intelligent and high-precision monitoring methods, this invention not only solves the problems of lagging quality monitoring and data silos in traditional cold chain transportation, but also provides the industry with a scalable technical framework, helping cold chain logistics upgrade to digitalization and standardization. BRIEF DESCRIPTION OF THE DRAWINGS

[0100] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0101] Figure 1 It is a calculation flow chart of the dynamic threshold generation of the present invention. DETAILED DESCRIPTION

[0102] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0103] Example 1:

[0104] This embodiment provides an intelligent early warning method for frozen shrimp transportation, comprising:

[0105] Step 1: Obtain real-time environmental data during the transportation of frozen shrimp through a third-party logistics system. The environmental data includes but is not limited to temperature, humidity, location, and acceleration data.

[0106] Step 2: The acquired data is transmitted to the data processing center, and the environmental data is analyzed and predicted in real time using deep learning networks (convolutional neural networks (CNN) and long short-term memory networks (LSTM), etc.). When the analysis and prediction results are greater than the preset threshold, the intelligent early warning mechanism is triggered, and relevant personnel are notified through various means, such as SMS, email, and mobile application push messages. In addition, the refrigeration, heating, ventilation and other equipment of the transport container can be automatically controlled and automatically adjusted to ensure the transportation environment of the frozen shrimp.

[0107] The preset thresholds include:

[0108] Set temperature ranges for different transportation stages, such as -18°C to -22°C for long-distance transportation;

[0109] Set a humidity limit to prevent high humidity from affecting the quality of freezing equipment or frozen shrimp. The upper limit is 80%;

[0110] Set acceleration thresholds for vibration and bumps. When the acceleration exceeds the threshold, an alarm is triggered to prevent packaging damage or frozen shrimp damage.

[0111] Combined with multi-dimensional data fusion technology, the threshold is dynamically adjusted to adapt to different transportation conditions and environmental changes.

[0112] The adaptive weight algorithm based on multi-dimensional data fusion includes the following steps:

[0113] (1) Data cleaning and time alignment.

[0114] (2) Construct the mutual information MINE network, which is expressed as:

[0115]

[0116] Among them, P XY (x,y) is the joint distribution; P X (x)P Y (y) is the product of the marginal distributions.

[0117] (3) Calculate the lower bound of the mutual information of each modal environment data, expressed as:

[0118]

[0119] Where: T θ (x,y) is a neural network, represents the calculation of the expectation of the joint sample, represents the expectation of computing edge samples (approximated by randomly shuffling x or y).

[0120] (4) Define the multidimensional feature quality vector, expressed as:

[0121] Q = [SNR, sampling rate, historical error rate, data integrity, timeliness, neighborhood consistency, environmental interference immunity, hardware health, signal entropy, drift coefficient, failure probability, calibration status]

[0122] The sources of these quality data are as follows:

[0123] SNR: Performs a Fourier transform (FFT) on the acquired signal to separate the signal and noise components in the frequency domain.

[0124] Sampling rate: The number of data points received within a statistical unit of time (such as 1 second).

[0125] Historical error rate: Aggregated calculation from a historical quality database (such as InfluxDB).

[0126] Data integrity: Checks whether data packets are lost (e.g., whether the MQTT message sequence numbers are continuous) and verifies whether data fields are complete (e.g., whether all key words in JSON exist).

[0127] Timeliness: Graded based on latency thresholds (e.g. <1ms = excellent, >100ms = poor).

[0128] Neighborhood consistency: Use clustering algorithms (K-means) or hypothesis tests (such as T-test) to determine whether the data deviates from the neighborhood mean.

[0129] Environmental interference immunity: Statistics of abnormal frequencies of sensors under specific environmental conditions.

[0130] Hardware health: Read sensor self-test status (such as SMART data and voltage fluctuations).

[0131] Signal entropy value: Calculate the Shannon entropy of the signal amplitude distribution:

[0132] Drift coefficient: For long-term trend analysis, the sliding window calculates the drift of the mean / variance (such as the linear regression slope).

[0133] Failure probability: Use classification algorithms (such as random forests) to predict the probability based on historical failure data.

[0134] Calibration status: record the last calibration time and results (such as whether the error is within the allowable range).

[0135] (5) Calculate the quality score.

[0136] First, calculate the dot product (weighted sum) of the multidimensional feature quality vector Q and the weight vector W:

[0137]

[0138] Among them, Q = [q1, q2, ..., q12] is a 12-dimensional quality feature vector (SNR, sampling rate, etc.), and W = [w1, w2, ..., w12] is the corresponding weight (obtained by AHP).

[0139] The original score S raw The final quality score is obtained by mapping it to the (0,1) interval through the Sigmoid function:

[0140]

[0141] Output range: S quality ∈(0,1) (the closer to 1, the higher the quality).

[0142] (6) Heterogeneous data collected by sensors Perform multimodal alignment calculations.

[0143] First, the mutual information of the two modal data is calculated through the constructed mutual information network. The calculation formula is:

[0144]

[0145] Among them, X i and X j are the sensor data of two modes respectively; T(x i ,x j ) represents the statistical function of the MINE network (neural network implementation), represents the joint distribution (positive sample pairs), Represents the marginal distribution product (negative sample pairs, by randomly hitting x i or x j generate).

[0146] Constructing a cross-modal attention matrix:

[0147]

[0148] Here, τ represents the temperature coefficient (hyperparameter).

[0149] Compute feature space projection:

[0150]

[0151] Among them, W i represents the private projection matrix of modality i, V j Represents the shared value transformation matrix of modality j.

[0152] (7) Dynamically weighted fusion of feature space projections is performed, and the calculation formula is:

[0153]

[0154] in, Represents the modal alignment feature, all modes are projected into a unified dimension d; quality_score (Q i ) represents the quality score of the quality vector of the i-th mode, expressed as s i .

[0155] Calculate dynamic weights and score the quality of each modality s i Perform exponential weighting + normalization (Softmax):

[0156]

[0157] The higher the quality of the modal, the greater the weight, and the sum of the weights is 1.

[0158] The weight w i Applied to aligned features Get the final fusion result Y:

[0159]

[0160] Y is consistent with the input feature time step and dimension.

[0161] (8) Threshold calculation is performed based on sliding window statistics (SPC method), and the calculation formula is:

[0162]

[0163] in: represents the mean value of Y in the time window [tw,t], represents the standard deviation of Y in the time window [tw,t], k represents the sensitivity coefficient (usually k = 2 to 3), and w represents the sliding window length (such as 60 time steps).

[0164] (9) Based on the comparison between Y and the threshold, different levels of warning are triggered.

[0165] normal:

[0166] if||Y t ||≤Thresholdt

[0167] warn:

[0168] if(Threshold t )<||Y t ||≤1.5·Threshold t

[0169] serious:

[0170] if||Y t ||>1.5·Threshold t

[0171] Example 2:

[0172] This embodiment provides an intelligent early warning system for frozen shrimp transportation, which uses the method described in Example 1 to provide early warning. The system includes:

[0173] The sensor module is used to collect environmental data during transportation, including temperature sensors, humidity sensors, position sensors, acceleration sensors, etc.

[0174] The data processing module uses a deep learning network to perform real-time analysis and prediction of the environmental data. When the analysis and prediction results are greater than a preset threshold, an intelligent early warning mechanism is triggered.

[0175] The preset thresholds include:

[0176] Set temperature ranges for different transportation stages, such as -18°C to -22°C for long-distance transportation;

[0177] Set a humidity limit to prevent high humidity from affecting the quality of freezing equipment or frozen shrimp. The upper limit is 80%;

[0178] Set acceleration thresholds for vibration and bumps. When the acceleration exceeds the threshold, an alarm is triggered to prevent packaging damage or frozen shrimp damage.

[0179] Combined with multi-dimensional data fusion technology, the threshold is dynamically adjusted to adapt to different transportation conditions and environmental changes.

[0180] The intelligent early warning mechanism includes: sending early warning information to management personnel through SMS, email, and mobile application push messages; automatically controlling the environmental control equipment of the transport container and automatically adjusting it to ensure the transportation environment of frozen shrimp.

[0181] The system of this embodiment also includes a user interface, which allows users to view real-time data, historical data, and warning records, and allows users to set and adjust preset thresholds.

[0182] Some steps in the embodiments of the present invention may be implemented using software, and the corresponding software program may be stored in a readable storage medium, such as a CD or a hard disk.

[0183] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent early warning method for frozen shrimp transportation, characterized in that: The method comprises: Step 1: Obtain environmental data during frozen shrimp transportation and perform preprocessing; Step 2: Use a deep learning network to perform real-time analysis and prediction of the environmental data. When the analysis and prediction results are greater than a preset threshold, an intelligent early warning mechanism is triggered. The analysis and forecasting process includes: (1) Constructing a mutual information network, which is represented by: Among them, P XY (x,y) is the joint distribution; P X (x)P Y (y) is the product of marginal distributions; (2) Calculate the lower bound of the mutual information of each modal environment data, expressed as: Among them, T θ (x,y) represents a neural network, represents the expectation of calculating the joint sample, Indicates the expectation of calculating marginal samples; (3) Define the multidimensional feature quality vector, expressed as: Q=[q1,q2,…q n ] Among them, q1,q2,…q n Indicates data quality parameters; (4) Calculate quality scores; First, the dot product of the multidimensional feature quality vector Q and the weight vector W is calculated: The weight vector W is obtained by AHP; The original score S raw The final quality score is obtained by mapping it to the (0,1) interval through the Sigmoid function: (5) Heterogeneous data collected by sensors Perform multimodal alignment calculation: First, the mutual information of the two modal data is calculated through the mutual information network. The calculation formula is: Among them, X i and Xj are the sensor data of two modes respectively; T(x i ,x j ) represents the statistical function of the mutual information network, represents the joint distribution, represents the marginal distribution product; Constructing a cross-modal attention matrix: Where τ represents the temperature coefficient; Compute feature space projection: Among them, W i represents the private projection matrix of modality i, V j represents the shared value transformation matrix of mode j; (6) Dynamically weighted fusion is performed on the feature space projection, and the calculation formula is: in, Represents the modal alignment feature, all modes are projected into a unified dimension d; quality_score (Q i ) represents the quality score of the i-th modal quality vector, denoted as si; Calculate dynamic weights and score the quality of each modality s i Perform exponential weighting + normalization: The weight w i Applied to aligned features Get the final fusion result Y: (7) Threshold calculation is performed based on the sliding window statistics method. The calculation formula is: in, represents the mean value of Y in the time window [tw,t], represents the standard deviation of Y in the time window [tw,t], k represents the sensitivity coefficient, and w represents the length of the sliding window; (8) The final fusion result Y and the threshold threshold t Make comparisons and trigger different levels of warnings.

2. The method according to claim 1, characterized in that The quality data in the multidimensional feature quality vector includes: SNR, sampling rate, historical error rate, data integrity, timeliness, neighborhood consistency, environmental interference resistance, hardware health, signal entropy value, drift coefficient, failure probability, and calibration status.

3. The method according to claim 1, characterized in that The environmental data includes: temperature, humidity, position, and acceleration.

4. The method according to claim 1, wherein The deep learning network includes: a convolutional neural network and a long short-term memory network.

5. An intelligent early warning system for frozen shrimp transportation, characterized in that: The system comprises: Sensor modules for collecting environmental data during transportation; A data processing module uses a deep learning network to perform real-time analysis and prediction of the environmental data. When the analysis and prediction results are greater than a preset threshold, an intelligent early warning mechanism is triggered; The analysis and forecasting process includes: (1) Constructing a mutual information network, which is represented by: Among them, P XY (x,y) is the joint distribution; P X (x)P Y (y) is the product of marginal distributions; (2) Calculate the lower bound of the mutual information of each modal environment data, expressed as: Among them, T θ (x,y) represents a neural network, represents the expectation of calculating the joint sample, Indicates the expectation of calculating marginal samples; (3) Define the multidimensional feature quality vector, expressed as: Q=[q1,q2,…q n ] Among them, q1,q2,…q n Indicates data quality parameters; (4) Calculate quality scores; First, the dot product of the multidimensional feature quality vector Q and the weight vector W is calculated: The weight vector W is obtained by AHP; The original score S raw The final quality score is obtained by mapping it to the (0,1) interval through the Sigmoid function: (5) Heterogeneous data collected by sensors Perform multimodal alignment calculation: First, the mutual information of the two modal data is calculated through the mutual information network. The calculation formula is: Among them, X i and Xj are the sensor data of two modes respectively; T(x i ,x j ) represents the statistical function of the mutual information network, represents the joint distribution, represents the marginal distribution product; Constructing a cross-modal attention matrix: Where τ represents the temperature coefficient; Compute feature space projection: Among them, W i represents the private projection matrix of modality i, V j represents the shared value transformation matrix of mode j; (6) Dynamically weighted fusion is performed on the feature space projection, and the calculation formula is: in, Represents the modal alignment feature, all modes are projected into a unified dimension d; quality_score (Q i ) represents the quality score of the i-th modal quality vector, denoted as si; Calculate dynamic weights and score the quality of each modality s i Perform exponential weighting + normalization: The weight w i Applied to aligned features Get the final fusion result Y: (7) Threshold calculation is performed based on the sliding window statistics method. The calculation formula is: in, represents the mean value of Y in the time window [tw,t], represents the standard deviation of Y in the time window [tw,t], k represents the sensitivity coefficient, and w represents the length of the sliding window; (8) The final fusion result Y and the threshold threshold t Make comparisons and trigger different levels of warnings.

6. The system according to claim 5, characterized in that The sensor module includes: a temperature sensor, a humidity sensor, a position sensor, and an acceleration sensor.

7. The system according to claim 5, characterized in that The intelligent early warning mechanism includes: Send early warning information to managers via SMS, email, and mobile app push messages; Automatically control the environmental control equipment of the transport container and automatically adjust to ensure the transportation environment of frozen shrimp.

8. The system according to claim 5, wherein: The preset thresholds also include: temperature thresholds, humidity thresholds, and acceleration thresholds for vibration and bumping set according to different transportation stages.

9. The system according to claim 5, characterized in that The system further includes a user interface.

10. The system according to claim 5, wherein: The data processing module uses a distributed storage system to store and manage environmental data.

Images