Epidemic wood whole-process intelligent supervision system and method based on block chain and Internet of Things

Abstract

The invention relates to the technical field of whole-process intelligent supervision, in particular to a whole-process intelligent supervision system and method for epidemic woods based on a block chain and the Internet of Things, and the system comprises an epidemic wood dynamic identification module, an integrated pressure sensor NFC chip is installed on the trunk of each epidemic wood, a pressure sensor is used for periodically collecting pressure values, monitoring the change of a packaging state, and carrying out the monitoring of the whole-process intelligent supervision of the epidemic woods. A new encryption sequence is generated through the NFC chip according to a preset time interval and stored, a mobile terminal APP reads the current encryption sequence of the NFC chip, the current encryption sequence and a background encryption key are matched and verified, and after verification is passed, a pressure value and the encryption sequence are combined. According to the method, the pressure sensor and the NFC chip are integrated for the trunk of each epidemic wood, real-time monitoring of packaging integrity and generation and verification of a dynamic encryption sequence are achieved, uniqueness and anti-counterfeiting capability of the identity of the source epidemic wood are improved, and it is ensured that each unit of epidemic wood can be accurately and safely tracked from the initial stage.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of whole-process intelligent supervision, and in particular to a whole-process intelligent supervision system and method for epidemic wood based on a blockchain and an Internet of Things. BACKGROUND

[0002] The technical field of whole-process intelligent supervision refers to the comprehensive use of new-generation information technologies such as an Internet of Things, big data, artificial intelligence, cloud computing and a blockchain to implement comprehensive monitoring and management of all links of a specific management object or business process, from source information collection, process real-time monitoring, data intelligent analysis to final decision feedback and closed-loop control.

[0003] In the specific application of the existing technical field of whole-process intelligent supervision, especially in the management of epidemic wood and other special goods, there are several inherent limitations. First, in terms of the fineness and real-time anti-counterfeiting of information collection, the commonly relied-upon data collection of an Internet of Things may not meet the demand for unique and dynamic encryption identification of each independent unit of epidemic wood, and traditional static two-dimensional codes or RFID tags are easy to be copied or replaced, making it difficult to effectively prevent source counterfeiting or midway switching, for example, part of a batch of epidemic wood is replaced with ordinary wood, and it is difficult to find out only by batch tracking. Second, although process real-time monitoring is mentioned, there is a lack of sufficient intelligent and automatic immediate intervention measures for complex situations that may occur during transportation, such as deviation from the predetermined route, abnormal parking, sudden weight changes and the like, and in most cases, it still remains in post-tracing, making it difficult to immediately stop the risk when it occurs, such as a vehicle deviating from the designated epidemic wood transportation route and dumping in a non-epidemic area. Therefore, improvements are needed. SUMMARY

[0004] The purpose of the present application is to solve the shortcomings in the prior art and to provide a whole-process intelligent supervision system and method for epidemic wood based on a blockchain and an Internet of Things.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solution: the whole-process intelligent supervision system for epidemic wood based on a blockchain and an Internet of Things comprises: an epidemic wood dynamic identification module, an NFC chip integrated with a pressure sensor is installed on the trunk of each epidemic wood, the pressure sensor is used to periodically collect pressure values to monitor the change of the packaging state, the NFC chip generates a new encrypted sequence at a preset time interval and stores it, the current encrypted sequence of the NFC chip is read by a mobile terminal APP, and is matched and verified with a background encryption key, after verification, the pressure values and the encrypted sequence are combined to generate epidemic wood encrypted identification information; The transportation intelligent monitoring module is deployed based on a vehicle-mounted three-in-one terminal, combines the encrypted identification information of the epidemic wood, synchronously collects load sensor readings, GPS trajectory data and 360° image key frame metadata, performs denoising and smoothing on the load sensor readings, performs drift correction on the GPS trajectory data, constructs a state vector including vehicle position, speed, acceleration, load and preset route compliance degree, dynamically adjusts state update parameters according to the current state vector and observation data residual error, predicts the vehicle position and load at the next moment, automatically compares the driving track in combination with the electronic fence and the preset transportation route, determines the state of abnormal deviation of more than 500 meters and records it, and obtains comprehensive data of the transportation process.

[0006] Preferably, the system further comprises: The on-chain data verification module extracts the load sensor readings at the loading time as the loading weight data based on the comprehensive data of the transportation process, intercepts the image data at the start and end stages of transportation as the transportation image data, records the loading weight data and the transportation image data in the distributed ledger of the blockchain platform after processing by an encryption algorithm, sets the forestry department node, the transportation enterprise node and the processing plant node to jointly maintain and access the distributed ledger, and establishes an epidemic wood storage electronic account book; The epidemic wood verification and verification module retrieves the unloaded wood image based on the epidemic wood storage electronic account book, extracts color, texture and size features, compares the feature vectors with the corresponding batch loading image data in the epidemic wood storage electronic account book, calculates the cosine similarity between the feature vectors as the matching degree, compares the matching degree with the preset threshold of 95%, reads the weighbridge weight data, calculates the error percentage by subtracting the loading weight data recorded in the epidemic wood storage electronic account book, compares the error percentage with the preset threshold of 3%, and determines whether to start the manual review process according to the comparison results, and obtains the epidemic wood verification and verification voucher.

[0007] Preferably, the epidemic wood dynamic identification module comprises: The pressure data acquisition sub-module uses a pressure sensor to continuously collect the pressure change corresponding to the electrical signal of the packaging layer of the epidemic wood trunk at a fixed sampling frequency, converts the electrical signal into a digital pressure value, monitors whether the pressure value difference between adjacent sampling points exceeds the preset fluctuation threshold to determine the packaging state change, and generates a packaging pressure sequence by aggregating the continuous pressure values; The encryption sequence generation sub-module calculates and generates a new encryption sequence according to a preset time period signal trigger, overwrites the new encryption sequence to the internal secure storage area of the chip, and obtains the NFC encryption code; The identification information combination sub-module, the mobile terminal APP reads the chip security storage area to obtain the NFC encryption code through the NFC communication protocol, sends the NFC encryption code to the background server, the server uses the pre-stored symmetric key to decrypt and check the sequence number of the NFC encryption code, and after checking, the latest pressure in the current packing pressure sequence and the NFC encryption code that passes the verification are structurally combined to generate the encrypted identification information of the vaccine wood.

[0008] Preferably, the transportation intelligent monitoring module comprises: The multi-source data acquisition sub-module unifies and aligns the timestamps of each sensor, synchronously acquires the output voltage signal of the load sensor to convert into a load reading, the longitude and latitude and speed data output by the GPS module, and the number and type of objects around the vehicle extracted from the interval photographed images of the 360° image acquisition device as key frame metadata, and obtains an original multi-source data packet; The state parameter correction sub-module, in combination with the encrypted identification information of the vaccine wood, removes noise from the load sensor reading in the original multi-source data packet using a sliding average filter, corrects the positioning drift of the GPS trajectory data using map data, constructs a state parameter set including the corrected vehicle position coordinates, instantaneous speed, tangential acceleration, smoothed load value, and distance from the current position to the preset electronic fence boundary as the preset route compliance state parameter, and obtains a corrected state parameter set; The trajectory deviation determination sub-module dynamically adjusts the diagonal elements of the process noise covariance matrix and the observation noise covariance matrix based on the differences between each state parameter in the corrected state parameter set and the predicted value at the previous time, predicts the vehicle position and load range within the future target time window, determines whether the predicted position continuously exceeds the 500-meter buffer zone in combination with the electronic fence geometric boundary and the preset transportation route center line, records the state, and obtains comprehensive transportation data.

[0009] Preferably, the on-chain data verification module comprises: The key information extraction sub-module filters a time point when the vehicle is stationary and the load is stable as the loading completion time based on the comprehensive transportation data, extracts the load sensor reading at that time as the loading weight data, intercepts the 360° image data frames within the target time period before and after the vehicle starts and stops as transportation image data, and establishes a set of key information to be stored; The data encryption and on-chain sub-module performs hash operation on the loading weight data and transportation image data in the set of key information to be stored to generate a digest, encrypts the digest using a preset asymmetric encryption public key, packages the encrypted digest and the original data copy together into a transaction data unit, submits it to a blockchain platform node for broadcasting and waits for block confirmation, and generates a block encryption evidence body.

[0010] Preferably, the on-chain data verification module further comprises a multi-party consensus maintenance submodule, wherein the forestry department node, the transportation enterprise node and the processing plant node are all set as blockchain network verification nodes, a preset consensus mechanism is used to verify and vote on the new block containing the block encryption evidence body, each node is set to have read and write permissions for different data fields in the distributed ledger, and an electronic ledger of the wood evidence is established.

[0011] Preferably, the wood evidence verification module comprises: An image feature comparison submodule, based on the electronic ledger of the wood evidence, the image of the wood pile is collected by the fixed camera at the unloading point, the contour of a single wood is extracted by image segmentation, the color histogram, the local binary pattern texture descriptor and the size of the circumscribed rectangle are calculated as features, the cosine similarity between the feature vectors of the two is calculated as the matching degree after the same type of features are extracted from the corresponding loading image data of the wood evidence electronic ledger and represented by vectors, and the image comparison similarity score is obtained. A weight difference verification submodule, the digital weighing data of the unloading site scale is read, the loading weight data of the corresponding transportation batch record is queried from the electronic ledger of the wood evidence, the weight difference is calculated, and the error percentage is calculated by dividing the weight difference by the loading weight data and multiplying by 100%.

[0012] Preferably, the wood evidence verification module further comprises a comprehensive review and determination submodule, wherein the image comparison similarity score is compared with a preset threshold of 95% to be greater than or equal to, and the absolute value of the weight verification difference rate is compared with a preset threshold of 3% to be less than or equal to, when any comparison result does not meet the condition, the batch of wood needs manual intervention inspection, and the wood evidence verification voucher is obtained.

[0013] The application also provides a wood whole-process intelligent supervision method, comprising the following steps: An integrated pressure sensor NFC chip is installed on the trunk of each wood, the pressure sensor is used to periodically collect pressure values, the packaging state change is monitored, the NFC chip is used to generate a new encryption sequence at a preset time interval and store it, the mobile terminal APP reads the current encryption sequence of the NFC chip, and matches and verifies it with the background encryption key, the pressure value and the encryption sequence are combined after verification to generate wood encryption identification information. The transportation intelligent monitoring module is deployed based on a vehicle-mounted three-in-one terminal, combines the said plague wood encrypted identification information, synchronously collects load sensor readings, GPS trajectory data and 360° image key frame metadata, denoises and smooths the load sensor readings, performs drift correction on the GPS trajectory data, constructs a state vector containing vehicle position, speed, acceleration, load and preset route compliance degree, dynamically adjusts state update parameters according to the current state vector and observation data residual error, predicts the vehicle position and load at the next moment, automatically compares the driving track with the preset transportation route in combination with the electronic fence, judges the state of abnormal deviation over 500 meters and records it, and obtains comprehensive data of the transportation process. The on-chain data verification module extracts the load sensor readings at the loading time as the loading weight data based on the said comprehensive data of the transportation process, intercepts the image data at the start and end stages of transportation as the transportation image data, records the loading weight data and the transportation image data in the distributed ledger of the block chain platform after processing by an encryption algorithm, sets forestry department nodes, transportation enterprise nodes and processing plant nodes to jointly maintain and access the distributed ledger, and establishes a plague wood storage electronic account book. The plague wood verification and verification module retrieves the unloading point collection plague wood image based on the said plague wood storage electronic account book, extracts color, texture and size features, compares the feature vectors with the corresponding batch loading image data in the plague wood storage electronic account book, calculates the cosine similarity between the feature vectors as the matching degree, compares the matching degree with the preset threshold 95%, simultaneously reads the weighbridge weighing data, calculates the error percentage by subtracting the loading weight data recorded in the plague wood storage electronic account book, compares the error percentage with the preset threshold 3%, decides whether to start the manual review process according to the two comparison results, and obtains the plague wood verification and verification voucher.

[0014] Compared with the prior art, the advantages and positive effects of the present application are that: In the application, by integrating a pressure sensor and an NFC chip for each trunk of the epidemic wood, real-time monitoring of the packaging integrity and generation and verification of the dynamic encryption sequence are realized, the uniqueness and anti-counterfeiting capability of the source epidemic wood identity are improved, and each unit of epidemic wood can be accurately and safely tracked from the initial stage. The vehicle-mounted terminal integrated with the load sensing, GPS trajectory tracking and 360° image acquisition device is deployed in the transportation link, the automatic comparison of the electronic fence and the preset route is combined, and the mechanism of triggering the vehicle locking when the abnormal deviation is more than 500 meters is set, the transparency and controllability of the transportation process are enhanced, the occurrence of behaviors such as stealing and illegal dumping is effectively prevented, and the traditional passive tracing is changed to active real-time intervention. The key data such as the loading weight and the transportation image are stored in the blockchain platform in real time, and the unalterable electronic account book is maintained by the forestry department, the transportation enterprise and the processing plant, a multi-party trust mechanism is constructed, the authenticity, integrity and non-repudiation of the data record are ensured, and a solid basis is provided for subsequent verification. The AI visual recognition technology is used to automatically compare the unloading epidemic wood features and the loading image data in the processing plant verification stage, and the double verification of the weighbridge weighing data is supplemented, the automatic and intelligent verification of the epidemic wood identity and quantity is realized, the verification efficiency and accuracy are improved, the human operation error and the fraud space are reduced, a whole closed-loop intelligent management from the source identification, the middle monitoring to the end verification is formed, and the epidemic wood loss risk is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The system flowchart of the application. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical scheme and advantages of the application clearer, the application is further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application, and do not limit the application.

[0017] Please refer to Figure 1 The application provides a technical scheme: an epidemic wood whole-process intelligent supervision system based on blockchain and Internet of Things, which comprises: An epidemic wood dynamic identification module, a pressure sensor NFC chip is installed and integrated on each trunk of the epidemic wood, the pressure sensor is used to periodically collect pressure values, the packaging state change is monitored, a new encryption sequence is generated by the NFC chip at a preset time interval and stored, the current encryption sequence of the NFC chip is read by a mobile terminal APP, and is matched and verified with a background encryption key, the pressure values and the encryption sequence are combined after the verification, and epidemic wood encryption identification information is generated; The transportation intelligent monitoring module is deployed based on a vehicle-mounted three-in-one terminal, combined with encrypted identification information of the epidemic wood, synchronously collects readings of a load sensor, GPS track data and 360° image key frame metadata, performs denoising and smoothing on the readings of the load sensor, performs drift correction on the GPS track data, constructs a state vector including a vehicle position, a speed, an acceleration, a load and a preset route compliance degree, dynamically adjusts state update parameters according to a current state vector and a residual error of observation data, predicts a vehicle position and a load at a next time, automatically compares a driving track with a preset transportation route in combination with an electronic fence, judges an abnormal deviation of more than 500 meters and records the same, and obtains comprehensive data in a transportation process; The on-chain data verification module extracts readings of the load sensor at a loading time as loading weight data, intercepts image data at a transportation start and end stage as transportation image data, records the loading weight data and the transportation image data in a distributed ledger of a blockchain platform after processing by an encryption algorithm, sets forestry department nodes, transportation enterprise nodes and processing plant nodes to jointly maintain and access the distributed ledger, and establishes an epidemic wood storage electronic account book; The epidemic wood verification module retrieves collected epidemic wood images at an unloading point, extracts color, texture and size features, compares feature vectors with corresponding batch loading image data in the epidemic wood storage electronic account book, calculates a cosine similarity between the feature vectors as a matching degree, compares the matching degree with a preset threshold of 95%, reads scale weighing data, calculates an error percentage by subtracting the loading weight data recorded in the epidemic wood storage electronic account book, compares the error percentage with a preset threshold of 3%, and determines whether to start an artificial review process according to the two comparison results to obtain an epidemic wood verification voucher.

[0018] The epidemic wood dynamic identification module includes: The pressure data acquisition sub-module continuously collects pressure changes of the packaging layer of the epidemic wood trunk corresponding to electrical signals by using a pressure sensor and setting a fixed sampling frequency, converts the electrical signals into digital pressure values, judges a packaging state change by monitoring whether a pressure value difference of adjacent sampling points exceeds a preset fluctuation threshold, and generates a packaging pressure sequence by aggregating continuous pressure values; The encryption sequence generation sub-module calculates and generates a new encryption sequence according to a preset time period signal, overwrites the new encryption sequence to a secure storage area in the chip, and obtains an NFC encryption code; The identification information combination sub-module reads the NFC encryption code from the secure storage area of the chip by using an NFC communication protocol, sends the NFC encryption code to a background server, decrypts and checks the sequence number of the NFC encryption code by using a pre-stored symmetric key, and combines the latest pressure in the current packaging pressure sequence and the verified NFC encryption code in a structure after the checking, to generate the epidemic wood encrypted identification information.

[0019] Specifically, a pressure sensor is used, which is fixed in advance in the packing layer area of the trunk, and its fixed sampling frequency is set to 5 times per second, i.e. the sampling period is 200 milliseconds. This frequency is selected according to the typical response time of the packing belt tightness change and the data processing ability of the NFC chip. The analog electrical signal corresponding to the pressure generated by the packing belt or rope on the trunk in the packing state is continuously collected, such as a voltage signal of 0 to 5 volts. This electrical signal is then converted by a 12-bit analog-to-digital converter (ADC) built into the sensor module or a separate ADC into a digital pressure raw reading ranging from 0 to 4095, and the raw reading is converted into a standard digital pressure value in kilopascals according to the sensor's calibration curve (e.g. 10 kilopascals per volt). Then, the difference between the digital pressure value obtained at the current sampling point and the digital pressure value obtained at the previous sampling point is calculated, and the absolute value of this difference is compared with a preset fluctuation threshold. This preset fluctuation threshold is obtained by statistical analysis of the pressure data of normal packed wood, specifically: collect at least 100 sets of pressure data of normal packed wood in a static state for one hour, calculate the average value and standard deviation of the pressure difference between adjacent sampling points in each set of data, and take the average of all group average values plus three times the average of the standard deviation as the reference. For example, if the average pressure difference between adjacent sampling points is 0.05 kilopascals and the average standard deviation is 0.02 kilopascals, the preset fluctuation threshold is set to 0.05 + 3 x 0.02 = 0.11 kilopascals. When the absolute value of the monitored pressure difference exceeds this 0.11 kilopascal threshold, the system determines that the packing state has changed significantly, which may indicate abnormal conditions such as packing belt loosening or breaking. Finally, the system collects and converts the digital pressure values and their corresponding time stamps collected over a period of time (e.g. the last 60 seconds) to form an ordered pressure data sequence, generating a packing pressure sequence.

[0020] According to one preset time period signal trigger operation, the time period signal is provided by the internal timer or external clock source of NFC chip, for example, set to trigger once every 10 minutes, this cycle takes into account the security requirements and chip power consumption, in order to prevent the risk of encryption sequence is sniffed for a long time to crack, when the trigger signal arrives, the internal microcontroller unit (MCU) of NFC chip executes the preset encryption algorithm program, the encryption algorithm is a pseudo-random number generation algorithm based on the current timestamp, chip unique serial number (UID) and an internal key (the key is written securely when the chip is manufactured or initialized, external unreadable) or a specific round function operation of a lightweight block cipher algorithm (such as PRESENT or SPECK), calculate through these input data, for example, splice the current Unix timestamp (second level) with the chip UID, then use the internal key to process the spliced data through the encryption mode operation of AES-128 algorithm, take part of its output (such as the first 16 bytes) as the new original sequence, and then perform a hash operation (such as SHA-256) on the original sequence, take the first several bytes (such as 8 bytes or 16 bytes) of the hash result as the final encryption sequence, ensure that each generated sequence has unpredictability and uniqueness, so as to generate a new encryption sequence, then the NFC chip overwrites the newly generated encryption sequence to the specified secure storage area in the chip, which usually has hardware level write protection and access control mechanism, to ensure that only authorized internal processes can modify, so as to obtain the NFC encryption code stored in the NFC chip.

[0021] The mobile APP provides an NFC reading function on its operation interface. When the user places the mobile device (such as a smart phone or a special handheld terminal, which needs to have an NFC reading and writing function) equipped with the APP close to the vaccine trunk with an integrated pressure sensor NFC chip, the NFC antenna of the mobile device and the chip antenna establish a near-field communication connection. The APP sends a reading instruction to the NFC chip through a standard NFC communication protocol (such as ISO / IEC14443 or ISO / IEC15693) to specify reading the NFC encryption code stored in the secure storage area inside the chip. After obtaining the NFC encryption code, the mobile APP immediately sends the NFC encryption code and some auxiliary information (such as device ID, current GPS location) of the current device to the designated API interface of the background server through a wireless network (such as 4G / 5G or Wi-Fi). After receiving the NFC encryption code, the background server first retrieves the pre-stored symmetric key of the batch or the area NFC chip from its secure key library. This symmetric key is the same as the key used to verify the serial number inside the NFC chip, or the server has a key for decrypting a specific encryption method. The server then performs decryption operations (if it is a symmetric encryption generated) or validity checks (for example, if the encryption code is generated based on a time synchronization algorithm such as TOTP, the server will generate a series of valid codes using the shared key and the current time window for comparison, or check whether the serial number contained inside is within the expected range and has not been replayed) on the received NFC encryption code using the pre-stored symmetric key. Through the verification, for example, confirming that the decrypted serial number is incremental and has not been used, or the TOTP verification passes, it is considered that the NFC encryption code is valid. Subsequently, the server combines the latest one or more pressure values in the packaged pressure sequence corresponding to the vaccine wood (this sequence is continuously generated by the pressure data acquisition submodule and may have been uploaded or will be uploaded together with the APP) with the NFC encryption code that has just passed the validation (or its validation success credentials, such as a session token) to form a structured combination, for example, into a JSON object, format {“pressure_latest”: 98.5, “pressure_sequence_preview”: [98.2, 98.3, 98.5], “nfc_code_validated”: “ABC123XYZ789”}, generating vaccine wood encryption identification information.

[0022] The transportation intelligent monitoring module comprises: The multi-source data acquisition submodule unifies and aligns the timestamps of each sensor, synchronously acquires the load sensor output voltage signal converted into load readings, the GPS module output latitude and longitude and speed data, and the number and type of objects around the vehicle extracted from the interval images of the 360° image acquisition device as key frame metadata, and obtains the original multi-source data packet. The state parameter correction sub-module combines the encrypted identification information of the vaccine wood and uses a sliding average filter to remove noise from the original multi-source data packet of the load sensor reading, applies map data to correct the positioning drift of the GPS trajectory data, constructs a state parameter set including the corrected vehicle position coordinates, instantaneous speed, tangential acceleration, smoothed load value, and the distance between the current position and the preset electronic fence boundary as the preset route compliance state parameter, and obtains the corrected state parameter set; The trajectory deviation determination sub-module dynamically adjusts the diagonal elements of the process noise covariance matrix and the observation noise covariance matrix based on the difference between each state parameter in the corrected state parameter set and the predicted value at the previous time, predicts the vehicle position and load range within the future target time window, combines the electronic fence geometric boundary with the center line of the preset transportation route, determines whether the predicted position continuously exceeds the 500-meter buffer zone, records the state, and obtains the comprehensive data of the transportation process.

[0023] Specifically, the timestamps of the data collected by each sensor (load sensor, GPS module, 360° image acquisition device) integrated in the vehicle-mounted three-in-one terminal are uniformly aligned and calibrated, for example, the UTC time output by the GPS module is used as the reference time source, the system clock of the vehicle-mounted terminal is synchronized through the network time protocol (NTP) or GPS time signal, and the timestamp error of all sensor data is ensured within a predetermined small range, for example, within 100 milliseconds, on this basis, the voltage signal output by the load sensor is synchronously collected, the voltage signal (for example, 0-10V) is converted into digital quantity through an analog-to-digital converter, and then converted into load reading according to the calibration parameters of the sensor (for example, 500 kilograms per volt), the unit is kilogram, the longitude and latitude coordinate data and instantaneous speed data (unit: kilometers / hour) output by the GPS module are collected, and in the image frames shot by the 360° image acquisition device at a predetermined interval (for example, every 10 seconds or every 100 meters of driving), the key visual elements of the vehicle's surrounding environment are extracted through a lightweight image analysis algorithm (for example, based on edge detection and contour analysis, or a pre-trained YOLOv_tiny target detection model), specifically, specific types of objects in the vehicle's field of view, such as the number of other vehicles, the number of pedestrians, and the approximate outline or number of large obstacles (such as buildings and trees), are identified and counted, these information is encoded as key frame metadata, for example, {“front_vehicles”: 2, “left_buildings”: 1, “timestamp”: “YYYY-MM-DDTHH:MM:SSZ”}, the above-mentioned synchronously collected load reading, GPS longitude and latitude and speed data, and 360° image key frame metadata are packaged together to form an original multi-source data packet containing a timestamp, and the original multi-source data packet is obtained.

[0024] In combination with the encryption identification information of the wood generated in the preceding step (mainly used for associating data and ensuring the validity of the data source, which is directly used here and does not participate in numerical calculation), each item of data in the original multi-source data packet is processed. Specifically, for the load sensor readings, a sliding average filtering method is used, with a sliding window size of 5 sampling points, i.e. the load readings of the current point and the previous 4 sampling points are taken to calculate the arithmetic mean value as the smoothed load value at the current time, to remove short-term fluctuations and sensor noise. For GPS trajectory data, first match the original latitude and longitude coordinate points with the preloaded high-precision electronic map data (e.g. a vector map accurate to the road level), and through a map matching algorithm (such as a matching based on a Hidden Markov Model HMM or a simple point-to-road segment projection algorithm), the drifting GPS points are adsorbed to the nearest road to correct the positioning drift and obtain the corrected vehicle position coordinates. Subsequently, based on the corrected GPS position coordinate sequence and the corresponding time stamp, the instantaneous speed (if the speed directly output by the GPS does not meet the accuracy requirement, then the displacement between adjacent position points and the time difference are used to calculate) and the tangential acceleration (the change in speed and the time difference are used to calculate) of the vehicle are calculated, and a state parameter set is constructed, which includes the corrected vehicle position coordinates (longitude, latitude), the calculated instantaneous speed, the calculated tangential acceleration, the smoothed load value after sliding average filtering, and the vertical distance from the current vehicle position to the nearest boundary of the preset electronic fence (the electronic fence is set by the forestry department according to the legal area of the wood transportation, such as the transportation starting point, the ending point and the polygon geographical boundary of the specific area along the way), which is one of the preset route compliance state parameters (if the distance is negative or exceeds a certain range, it may deviate from the preset area), to obtain the corrected state parameter set.

[0025] The difference between each state parameter in the corrected state parameter set obtained at the last time (corrected vehicle position coordinates, instantaneous speed, tangential acceleration, smoothed load value, current position and distance from the preset electronic fence boundary) and the predicted value output by the Kalman filter or its variants (such as extended Kalman filter EKF) prediction model at the previous time (or multiple historical times) is used to dynamically adjust the diagonal elements of the process noise covariance matrix Q and the observation noise covariance matrix R in the Kalman filter model, for example, if the observation residual of the position is continuously large, the variance of the corresponding position state in the process noise covariance matrix Q is appropriately increased, and the variance of the corresponding position observation in the observation noise covariance matrix R is appropriately reduced, so that the filter relies more on the actual observation value, and vice versa. This adjustment is based on a preset rule, for example, when the residual exceeds 2 times its expected standard deviation (estimated according to historical data or model parameters), adjust the corresponding noise term by 10% of its current value. By adjusting the filter parameters, the corrected state parameter set at the current time is used as the new observation input to predict the possible position range (a confidence ellipse or rectangular region centered on the prediction point, combined with the prediction covariance) and load range (prediction value plus or minus three times the prediction standard deviation) of the vehicle within a target time window in the future (for example, 1 to 5 minutes in the future). Then, the predicted vehicle position range is compared with the preset electronic fence geometric boundary and the preset transportation route (usually a center line composed of a series of geographic coordinate points) to determine whether the predicted vehicle position is continuously (for example, for 3 consecutive prediction time steps, each step being 1 minute) outside the 500-meter buffer zone on both sides of the preset transportation route center line. The 500-meter buffer zone is set based on the experience evaluation of the reasonable detour distance of the transportation vehicle due to traffic conditions, road construction, etc., and is determined comprehensively considering the GPS positioning accuracy. If it is determined that it is continuously outside the 500-meter buffer zone, the state (including time, position, deviation distance, etc.) is recorded as an abnormal deviation event. Comprehensive data of the transportation process is obtained by integrating all collected, corrected, predicted and judged data.

[0026] The on-chain data verification module includes: The key information extraction submodule filters the time points when the vehicle is stationary and the load is stable as the loading completion time based on the transportation process comprehensive data, extracts the load sensor reading at that time as the loading weight data, intercepts the 360° image data frames in the target time period before and after the vehicle starts as the transportation image data, and establishes a set of key information to be stored. The data encryption on-chain submodule generates an abstract by performing a hash operation on the key information to be stored, which includes the loading weight data and the transportation image data, respectively, encrypts the abstract using a preset asymmetric encryption public key, packs the encrypted abstract and the original data copy together into a transaction data unit, submits it to the blockchain platform node for broadcasting and waits for block confirmation, and generates a block encryption evidence body; The on-chain data verification module further includes a multi-party consensus maintenance submodule, which sets the forestry department node, the transportation enterprise node and the processing plant node as blockchain network verification nodes, verifies and votes on the new block containing the block encryption evidence body through a preset consensus mechanism, sets the read and write permissions of each node for different data fields in the distributed ledger, and establishes an electronic ledger for forest evidence.

[0027] Specifically, based on the comprehensive transportation data, which includes vehicle GPS trajectory, speed, load sensor reading sequence and 360° image key frame metadata, first, the vehicle GPS speed data is analyzed, and the judgment standard for vehicle static is set as the speed reading being lower than 0.5 kilometers per hour within 180 seconds. The threshold of 0.5 kilometers per hour is set according to the typical drift error range of GPS in low speed or static state and the actual loading operation condition that the vehicle basically does not travel. The time length of 180 seconds is used to exclude the case of temporary stop such as waiting for traffic signal. Then, within the identified vehicle static time period, the stability of the load sensor reading sequence is further analyzed, the standard deviation of the load reading in each continuous 30-second window within this time period is calculated, and if the standard deviation is less than 15 kilograms, it is considered that the load is stable within the 30-second window. The threshold of 15 kilograms is set based on the accuracy of the load sensor and the minimum significant change in weight when loading a single or small batch of forest, and the time sequence of the last sampling time that meets the load stability condition is selected as the loading completion time. Then, the smoothed load sensor reading recorded at the loading completion time is extracted as the loading weight data, for example, 15250.5 kilograms. Subsequently, all image data frames recorded by the 360° image acquisition device within the target time period from 300 seconds before the loading completion time to 60 seconds after the loading completion time are selected as the transportation image data, or key frames are selected according to the strategy of selecting one frame every 15 seconds, to establish the key information set to be stored.

[0028] The truckload weight data and transportation image data (or metadata list) in the key information set to be stored in the previous step are subjected to hash operation respectively. Specifically, for the truckload weight data accurate to the decimal point, such as "15250.5", after being converted into a standard string format, the SHA-256 algorithm is used to calculate its hash digest, obtaining a 64-bit hexadecimal string. For transportation image data, if it is multiple frames of images, first calculate the hash value of each image file by SHA-256 algorithm, then splice these hash values into a string in chronological order, and then use SHA-256 algorithm to calculate the final image data hash digest of the spliced total string, or if the image data is stored in IPFS or other distributed storage, the content identifier (CID) is directly used as the content to be hashed or the hash digest itself. Subsequently, a preset asymmetric encryption public key, for example, a 2048-bit RSA public key issued and managed by the forestry administrative department, is used to encrypt the truckload weight data hash digest and transportation image data hash digest generated in the foregoing by an asymmetric encryption algorithm (such as RSA-OAEP), to generate encrypted digests. After that, the two encrypted digests are organized into a structured transaction data unit together with the original truckload weight data (for example, "15250.5 kg") and a copy of the transportation image data (if it is an original image file, the copy here usually refers to the storage path or IPFS address of the image file, as well as the thumbnail or metadata of the key frame; if the original data is small, such as only a metadata list, it is directly included), such as a JSON object, which contains fields such as "encryptedWeightHash", "encryptedImageHash", "originalWeight", "imageReferences", and is accompanied by time stamp, vehicle identification and other related information. The transaction data unit is submitted to the predetermined configured blockchain platform node through the API interface, the node receives and performs local verification, and broadcasts the transaction to the whole network through the P2P network, waits for the transaction to be packaged into a new block and pass through the consensus mechanism (for example, waits for at least 6 block confirmations, this number is set based on the experience probability of the finality of the blockchain, such as on the similar public chain of Ethereum, or waits for a specified number of endorsement nodes to sign and ordering nodes to confirm in the consortium chain), and generates a block encryption notarization body.

[0029] The nodes of the forestry department, the nodes of the transportation company, and the nodes of the processing plant are all set as verification nodes of the blockchain network, which means that the operators of these nodes are granted the right to participate in the verification and consensus process during the initialization or subsequent governance of the blockchain network. For example, in a Hyperledger Fabric-based consortium chain, the nodes of these organizations are configured as part of the endorsement nodes or ordering nodes, and their identity and authority are identified through digital certificates managed by the Member Service Provider (MSP). The system uses a pre-set consensus mechanism, such as Raft or IBFT majority voting or Byzantine Fault Tolerance protocol, to verify the validity of new block proposals broadcast by other nodes in the network, which contain encrypted evidence of newly generated blocks. The verification content includes checking the correctness of the transaction format, the validity of the digital signature, whether the transaction content meets the business logic specified by the smart contract (e.g., whether the loading weight is within a reasonable range, whether the associated vehicle and batch of logs have been registered, etc.), and whether the pre-set endorsement policy is met (e.g., a new log loading record on-chain transaction may require at least the joint endorsement signature of the forestry department node and the transportation company node). In the voting stage, each verification node decides whether to agree to add the block to its own copy of the ledger based on the verification result and broadcasts its vote. When the required number of votes (e.g., more than two-thirds of the verification nodes agree) is reached, the new block is officially confirmed and linked to the chain. At the same time, through the access control logic of the smart contract or chaincode, the read and write permissions of each type of node (forestry department, transportation company, processing plant) for different data fields in the distributed ledger are fine-tuned. For example, the forestry department node may have the highest read permission for all log supervision data and write permission for critical status (such as the cancellation status). The transportation company node can only read and write transportation process data and loading data related to its own transportation tasks, while the processing plant node mainly reads log information of the batch to be received and writes relevant data of log processing and cancellation. This permission setting is based on the principle of least privilege, and is implemented through conditional judgment statements such as require(msg.sender == transportCompanyAddress && batch.carrier == msg.sender) in the smart contract to establish an electronic log evidence account.

[0030] The log cancellation inspection module includes: The image feature comparison submodule, based on the electronic inventory of forest products, calls the image of the forest product stack collected by the fixed camera at the unloading point, extracts the contour of a single forest product through image segmentation, calculates the color histogram, local binary pattern texture descriptor, and the size of the circumscribed rectangle as features, and then, after the same features are extracted from the corresponding batch loading image data in the electronic inventory of forest products and are represented as vectors, the cosine similarity between the two feature vectors is calculated as the matching degree, and the image comparison similarity score is obtained; The weight difference verification submodule reads the digital weighing data output by the scale at the unloading site, queries the loading weight data of the corresponding transport batch from the electronic inventory of forest products, calculates the weight difference, and then calculates the error percentage by dividing the weight difference by the loading weight data and multiplying by 100. The forest product verification module also includes a comprehensive review and determination submodule that compares the image comparison similarity score with the preset threshold of 95% and compares the absolute value of the weight verification difference rate with the preset threshold of 3%. If either comparison result does not meet the conditions, the batch of forest products is marked for manual intervention inspection, and a forest product verification certificate is obtained.

[0031] Specifically, based on the electronic inventory of forest products established in the preceding steps, the 360° image data frames recorded during loading for the corresponding batch are first called from the electronic inventory by batch number or transport task number, or their feature vectors. At the same time, at the forest product unloading point, real-time collection of the current unloaded forest product stack image is performed using a preset fixed position high-definition camera to obtain the unloading point forest product image. Then, the forest product stack image collected at the unloading point is subjected to image segmentation processing to extract the clear contour of a single or small cluster of forest products. This image segmentation uses an instance segmentation model based on deep learning, such as MaskR-CNN. This model generates candidate regions through its region proposal network (RPN), and then processes them in parallel by the classification head, bounding box regression head, and mask head, outputting the class, bounding box, and pixel-level mask of each forest product instance. During model training, the input is an image containing forest products and its corresponding hand-labeled mask, and the output is a predicted mask. The cross-entropy loss function is used for mask prediction optimization, and the classification and bounding box regression loss are jointly trained. The Adam optimizer is used with an initial learning rate of 10 -3and decay with the training rounds until the performance on the validation set no longer improves significantly. The model structure usually contains a ResNet-50 or ResNet-101 as the backbone network to extract features, followed by a feature pyramid network (FPN) to fuse multi-scale features, and then a RPN and subsequent detection head and mask head. After segmenting the single tree trunk outline, for each segmented tree trunk region and the corresponding tree trunk region in the loading image data retrieved from the electronic account (if the loading image is not preprocessed, segmentation is also performed), the color histogram features are calculated, such as converting the image to HSV color space, calculating the normalized histogram of 32 intervals for each of the H, S, and V channels, and concatenating them into a 96-dimensional vector. The local binary pattern (LBP) texture descriptor is calculated, such as using a radius of 1 and a neighborhood of 8, and using a uniform pattern to reduce the LBP code histogram to 59 dimensions. The normalized width, height, and aspect ratio of the minimum bounding rectangle are calculated as size features (3 dimensions). These extracted color, texture, and size features are combined into a high-dimensional feature vector. After vectorizing the same type of feature vectors extracted from the loading image and unloading image, the cosine similarity between the two feature vector sets (e.g., taking the average of all single tree trunk feature vectors within each batch, or using more complex set matching methods) is calculated as the matching degree, and the formula is where V 装车 and V 卸货 are the average feature vectors extracted from the loading image and unloading image, respectively, and the image comparison similarity is obtained.

[0032] First, after the tree trunk transport vehicle arrives at the unloading site, the vehicle as a whole or only the part loaded with tree trunks (such as the cargo box) is driven onto the weighbridge. After the weighbridge reading stabilizes, the digital weighing data output by the weighbridge system is read, which is the actual unloading weight of the tree trunks at this unloading point, for example, recorded as 14980.5 kg. Then, according to the current batch number or transport task number of the tree trunks being processed, the loading weight data recorded and already chained and stored in the tree trunk evidence electronic account established and maintained in the previous step is retrieved through a query operation, for example, the query obtains a loading weight of 15250.5 kg. Next, the difference between the unloading weight data and the loading weight data is calculated, i.e., weight difference = loading weight data - unloading weight data. Using the above example, weight difference = 15250.5 kg - 14980.5 kg = 270.0 kg. Then, the calculated weight difference is divided by the original loading weight data, and the result is multiplied by 100 to obtain the percentage of weight difference, i.e., Using the above example, The weight verification difference rate is obtained.

[0033] The image comparison similarity score obtained in the previous step, for example 0.962, is compared with a pre-set matching degree threshold, which is set to 95% (i.e. 0.95). This threshold is set based on a large number of known matching and non-matching pairs of forest product loading and unloading image pairs, and the distribution of similarity scores is statistically analyzed to select a boundary point that can identify as many true matching pairs as possible (e.g. a recall rate of more than 98%) at an acceptable false positive rate (e.g. less than 2%), for example by drawing an ROC curve and selecting the best operating point to determine 0.95. A greater than or equal comparison is performed, i.e. whether the image comparison similarity score is greater than or equal to 0.95, and the absolute value of the weight verification difference rate obtained in the previous step, for example |1.770%| = 1.770%, is compared with a pre-set weight error percentage threshold, which is set to 3%. This threshold is set in reference to the forestry industry's standard for reasonable weight loss caused by natural factors such as water evaporation and transit consumption during bulk wood transportation, and combined with historical transportation data statistical analysis, for example, if the statistics show that 99% of the weight difference absolute values of compliant transportation batches are within 3%, then 3% can be used as the threshold. A less than or equal comparison is performed, i.e. whether the absolute value of the weight verification difference rate is less than or equal to 3%, based on the results of the two independent comparisons. When either of the conditions that the image comparison similarity score is less than 0.95 or the absolute value of the weight verification difference rate is greater than 3% is met, the system automatically marks the quarantine wood verification result of this batch as abnormal and triggers the manual intervention inspection process, prompting the management personnel to conduct detailed on-site verification and cause tracing on the batch of quarantine wood, and obtaining the quarantine wood verification certificate.

Claims

1. The intelligent supervision system for the whole process of diseased trees based on blockchain and the Internet of Things is characterized by: The system comprises: Diseased tree dynamic identification module, each diseased tree trunk is installed with an integrated pressure sensor NFC chip, the pressure sensor is used to periodically collect pressure values, monitor changes in packaging status, and generate and store new encryption sequences at preset time intervals through the NFC chip. The mobile terminal APP reads the current encryption sequence of the NFC chip and matches it with the background encryption key for verification. After verification, the pressure value and encryption sequence are combined to generate the encrypted identification information of the diseased tree; The intelligent transportation monitoring module is deployed based on the on-board three-in-one terminal and combined with the encrypted identification information of the disease tree, synchronously collects the load sensor readings, GPS trajectory data and 360° image key frame metadata, denoises and smoothes the load sensor readings, performs drift correction on the GPS trajectory data, and constructs a state vector including vehicle position, speed, acceleration, load and preset route compliance. It dynamically adjusts the state update parameters based on the current state vector and the residual of the observed data, predicts the vehicle position and load at the next moment, automatically compares the driving trajectory with the preset transportation route in combination with the electronic fence, determines and records the abnormal deviation of more than 500 meters, and obtains comprehensive data of the transportation process.

2. The intelligent supervision system for the whole process of diseased trees based on blockchain and the Internet of Things according to claim 1 is characterized in that: The system further comprises: The on-chain data verification module, based on the comprehensive data of the transportation process, extracts the load sensor reading at the loading time as the loading weight data, intercepts the image data of the transportation start and end stages as the transportation image data, processes the loading weight data and transportation image data through an encryption algorithm, and records them in the distributed ledger of the blockchain platform. The forestry department node, the transportation enterprise node, and the processing plant node are set to jointly maintain and control access rights of the distributed ledger, and establish an electronic ledger for evidence of diseased trees; The diseased wood verification and inspection module, based on the electronic ledger of diseased wood evidence, retrieves the diseased wood images collected at the unloading point, extracts color, texture and size features, compares the feature vectors with the corresponding batch loading image data in the electronic ledger of diseased wood evidence, calculates the cosine similarity between the feature vectors as the matching degree, and compares the matching degree with the preset threshold value of 95%. At the same time, the weighing data of the scale is read, and the error percentage is calculated by difference with the loading weight data recorded in the electronic ledger of diseased wood evidence, and the error percentage is compared with the preset threshold value of 3%. Based on the results of the two comparisons, it is decided whether to start the manual review process to obtain the diseased wood verification certificate.

3. The intelligent supervision system for the whole process of diseased trees based on blockchain and the Internet of Things according to claim 1 is characterized in that: The diseased tree dynamic identification module includes: The pressure data acquisition submodule uses a pressure sensor to set a fixed sampling frequency, continuously collects electrical signals corresponding to pressure changes in the packing layer of the infected tree trunk, converts the electrical signals into digital pressure values, monitors whether the pressure value difference between adjacent sampling points exceeds the preset fluctuation threshold to determine the change in the packing state, aggregates continuous pressure values, and generates a packing pressure sequence; The encryption sequence generation submodule is triggered by a preset time period signal, calculates and generates a new encryption sequence, overwrites the new encryption sequence to the chip's internal secure storage area, and obtains the NFC encryption code; The identification information combination submodule, the mobile APP reads the chip security storage area through the NFC communication protocol to obtain the NFC encryption code, sends the NFC encryption code to the background server, and the server uses the pre-stored symmetric key to decrypt the NFC encryption code and verify the serial number. After verification, the latest pressure in the current packaging pressure sequence and the verification NFC encryption code are structured and combined to generate encrypted identification information.

4. The intelligent supervision system for the whole process of diseased trees based on blockchain and the Internet of Things according to claim 1 is characterized in that: The transport intelligent monitoring module includes: The multi-source data acquisition submodule aligns the acquisition timestamps of each sensor, synchronously collects the load sensor output voltage signal, converts it into load readings, the GPS module outputs latitude, longitude and speed data, and extracts the number and type of objects around the vehicle from the interval-shot images of the 360° image acquisition device as key frame metadata to obtain the original multi-source data packet; The state parameter correction submodule, in combination with the encrypted identification information, uses a sliding average filter to remove noise from the load sensor readings in the original multi-source data packet, applies map data to the GPS trajectory data to correct positioning drift, and constructs a set of preset route compliance state parameters including the corrected vehicle position coordinates, instantaneous speed, tangential acceleration, smoothed load value, and the distance between the current position and the preset electronic fence boundary, to obtain a corrected state parameter set; The trajectory deviation determination submodule dynamically adjusts the diagonal elements of the process noise covariance matrix and the observation noise covariance matrix based on the difference between each state parameter in the corrected state parameter set and the predicted value at the previous moment, predicts the vehicle position and load range within the future target time window, and combines the geometric boundary of the electronic fence with the center line of the preset transportation route to determine whether the predicted position continues to exceed the 500-meter buffer zone and records this status to obtain comprehensive data on the transportation process.

5. The intelligent supervision system for the whole process of diseased trees based on blockchain and the Internet of Things according to claim 2 is characterized in that: The on-chain data verification module includes: The key information extraction submodule, based on the comprehensive data of the transportation process, selects the time point when the vehicle is stationary and the load is stable as the loading completion time point, extracts the load sensor reading at that time point as the loaded weight data, intercepts 360° image data frames within the target time period before and after the vehicle starts as transportation image data, and establishes a key information set to be stored; The data encryption and chain sub-module performs hash operations on the key information to be stored, including the loading weight data and the transportation image data, to generate a summary, encrypts the summary using the preset asymmetric encryption public key, and packages the encrypted summary and the original data copy into a transaction data unit. The unit is submitted to the blockchain platform node for broadcast and waits for block confirmation to generate a block encrypted evidence body.

6. The intelligent supervision system for the whole process of diseased trees based on blockchain and the Internet of Things according to claim 5 is characterized in that: The on-chain data verification module also includes: a multi-party consensus maintenance sub-module, which sets the forestry department node, transportation enterprise node and processing plant node as blockchain network verification nodes, verifies and votes on new blocks containing block encrypted evidence through a preset consensus mechanism, sets the reading and writing permissions of each node for different data fields in the distributed ledger, and establishes an electronic ledger of diseased wood evidence.

7. The intelligent supervision system for the whole process of diseased trees based on blockchain and the Internet of Things according to claim 2 is characterized in that: The diseased wood verification module includes: The image feature comparison submodule, based on the electronic record of diseased wood evidence, retrieves the fixed camera at the unloading point to collect images of the diseased wood pile, extracts the outline of a single diseased wood through image segmentation, calculates the color histogram, local binary pattern texture descriptor and circumscribed rectangle size as features, extracts similar features from the corresponding batch loading image data in the electronic record of diseased wood evidence, and vectorizes them. The cosine similarity between the feature vectors of the two is calculated as the matching degree to obtain the image comparison similarity score; The weight difference verification submodule reads the digital weighing data output by the scale at the unloading site, queries the loading weight data of the corresponding transport batch record from the electronic ledger of Yimu Evidence, calculates the weight difference, then divides the weight difference by the loading weight data and multiplies it by 100% to calculate the error percentage and obtain the weight verification difference rate.

8. The intelligent supervision system for the whole process of diseased trees based on blockchain and the Internet of Things according to claim 7 is characterized in that: The diseased wood verification and inspection module also includes: a comprehensive review and judgment sub-module, which compares the image comparison similarity score with a preset threshold of 95% to be greater than or equal to, and compares the absolute value of the weight verification difference rate with a preset threshold of 3% to be less than or equal to. When any comparison result does not meet the conditions, the batch of diseased wood is marked as requiring manual intervention and inspection to obtain a diseased wood verification certificate.

9. The method for intelligent supervision of diseased trees throughout the entire process based on the blockchain and Internet of Things according to any one of claims 1 to 8, characterized in that: The following steps are involved: An integrated pressure sensor NFC chip is installed on the trunk of each infected tree. The pressure sensor is used to periodically collect pressure values ​​to monitor changes in packaging status. A new encryption sequence is generated and stored through the NFC chip at preset time intervals. The mobile app reads the current encryption sequence of the NFC chip and matches it with the background encryption key. After verification, the pressure value and the encryption sequence are combined to generate encrypted identification information of the infected tree. The intelligent transportation monitoring module is deployed based on the on-board three-in-one terminal. In combination with the encrypted identification information of the disease tree, it synchronously collects load sensor readings, GPS trajectory data and 360° image key frame metadata, denoises and smoothes the load sensor readings, performs drift correction on the GPS trajectory data, and constructs a state vector containing vehicle position, speed, acceleration, load and preset route compliance. The state update parameters are dynamically adjusted according to the residual of the current state vector and the observed data, and the vehicle position and load at the next moment are predicted. The driving trajectory is automatically compared with the preset transportation route in combination with the electronic fence, and the abnormal deviation of more than 500 meters is determined and recorded to obtain comprehensive data of the transportation process; The on-chain data verification module, based on the comprehensive data of the transportation process, extracts the load sensor reading at the loading time as the loading weight data, intercepts the image data of the transportation start and end stages as the transportation image data, processes the loading weight data and transportation image data through an encryption algorithm, and records them in the distributed ledger of the blockchain platform. The forestry department node, the transportation enterprise node, and the processing plant node are set to jointly maintain and control access rights of the distributed ledger, and establish an electronic ledger for evidence of diseased trees; The diseased wood verification and inspection module, based on the electronic ledger of diseased wood evidence, retrieves the diseased wood images collected at the unloading point, extracts color, texture and size features, compares the feature vectors with the corresponding batch loading image data in the electronic ledger of diseased wood evidence, calculates the cosine similarity between the feature vectors as the matching degree, and compares the matching degree with the preset threshold value of 95%. At the same time, the weighing data of the scale is read, and the error percentage is calculated by difference with the loading weight data recorded in the electronic ledger of diseased wood evidence, and the error percentage is compared with the preset threshold value of 3%. Based on the results of the two comparisons, it is decided whether to start the manual review process to obtain the diseased wood verification certificate.

Images