Intelligent evidence obtaining method and system for delivery accident of courier, and electronic equipment

Abstract

The invention relates to the technical field of courier delivery accident intelligent evidence obtaining scheme design, in particular to a courier delivery accident intelligent evidence obtaining method and system and electronic equipment. The method comprises the following steps: detecting acceleration abrupt change through a distribution terminal IMU to trigger multi-source evidence obtaining; synchronously calling double cameras and a microphone to acquire a video stream, a depth map and an environment audio; analyzing courier actions and cargo tracks based on a space-time attention mechanism; detecting collision and noisy events through the audio spectrum; integrating IMU, video and audio data to construct a cause and effect event atlas with synchronous timestamps; utilizing the pre-training responsibility quantification model to output responsibility coefficients of the courier, the third party and the environmental factors; writing the encrypted data and the geo-fence information into a block chain; and generating an intelligent report containing the three-dimensional reconstruction animation and the responsibility matrix. According to the invention, multi-modal data automatic alignment and block chain evidence fixation are realized, and accident liability determination errors are reduced to a great extent.

Description

Technical Field

[0001] The present invention relates to the technical field of design of intelligent evidence collection schemes for courier delivery accidents, and in particular to an intelligent evidence collection method and system as well as electronic equipment for courier delivery accidents. Background Art

[0002] The current express delivery industry generally relies on manual backtracking to handle delivery accidents, which presents significant technical flaws. First, evidence collection methods are passive and lagging: traditional methods rely on post-event surveillance or user complaints. Key dynamic information at the moment of the accident (such as the trajectory of the cargo leaving the hands and the timing of the external force collision) is permanently lost due to the equipment not being activated in time. Second, the chain of evidence is fragmented: video, audio, and motion sensor data are stored independently and with misaligned timestamps, making it difficult to establish multimodal event correlations and, consequently, unable to trace third-party interference liability. Finally, liability determination is highly subjective: lacking automated analysis models, manual assessments are susceptible to empirical bias, particularly inaccurately estimating risk weights for environments such as slippery roads and slopes. While some solutions utilize continuous recording like dashcams, this continuous recording mode generates a large amount of redundant data, increasing terminal power consumption and making it difficult to retrieve valid accident footage. Furthermore, existing blockchain evidence storage is mostly limited to simple data on-chain storage and lacks integration with three-dimensional physical simulation, resulting in limited legal validity. Therefore, there is an urgent need to develop a forensic solution that can be proactively triggered, synchronized across multiple sources, and intelligently attributed to address the core pain points of insufficient evidence integrity and inaccurate liability quantification.

[0003] Therefore, the existing technology needs to be further developed. Summary of the Invention

[0004] The purpose of the present invention is to overcome the above-mentioned technical deficiencies and provide an intelligent evidence collection method and system and electronic equipment for courier delivery accidents to solve the problems existing in the prior art.

[0005] To achieve the above technical objectives, according to a first aspect of the present invention, the present invention provides an intelligent evidence collection method for courier delivery accidents, comprising: S100 monitors the courier's motion status in real time through the delivery terminal's built-in inertial measurement unit. When a sudden acceleration exceeding a preset threshold is detected, the multi-source evidence collection module is automatically activated, synchronously triggering the delivery terminal's front and rear cameras and microphone to collect multi-angle video streams, depth map data, and ambient audio from the accident scene. S200: Perform real-time frame sequence analysis on the video stream, identify the courier's body movements and the trajectory of the cargo drop through the spatiotemporal attention mechanism; detect key event points of collision sounds and human voice arguments based on the audio spectrum characteristics collected by the microphone; S300, cross-modally aligning IMU data, video motion features, and audio event points to construct a causal event graph with synchronized timestamps; calling a pre-trained accident responsibility quantification model to output the courier responsibility coefficient, third-party responsibility coefficient, and environmental factor weights based on the causal event graph; S400: Encrypt the multi-source forensic data and its timestamp and geo-fence information and write them into the blockchain distributed ledger; generate an intelligent forensic report including a three-dimensional scene reconstruction animation, a responsibility coefficient matrix, and a blockchain evidence storage number.

[0006] Specifically, the spatiotemporal attention mechanism includes: Track the key points of the human skeleton in the video frame sequence and calculate the rate of change of the courier's torso inclination; The YOLO model is used to detect the trajectory of the parabola and mark the coordinates of the collision area with the ground.

[0007] Specifically, the cross-modal alignment adopts a dynamic time warping algorithm, and the alignment method satisfies: The deviation between the peak moment of IMZ axial acceleration and the frame when the cargo leaves the hand in the video is controlled within ±100ms; The time synchronization error between the audio voiceprint mutation point and the video collision frame is less than 50% of the audio frame length.

[0008] Specifically, the training method of the accident responsibility quantification model includes: Use a generative adversarial network to synthesize 100,000 delivery accident scene samples; A weakly supervised learning framework is used to label the ternary responsibility labels of courier error, third-party collision, and slippery road surface.

[0009] Specifically, it also includes: When a human quarrel is detected, the voice emotion analysis module is activated to extract the anger index and threat keyword frequency in the voiceprint features, and generate a chain of evidence of provocative behavior by a third party.

[0010] Specifically, the speech emotion analysis adopts: Mel-frequency cepstral coefficients combined with dual-stream feature extraction of BERT speech encoder; Localizing the temporal intervals of abusive sentences via gradient class activation mapping.

[0011] Specifically, the generation of the 3D scene reconstruction animation includes: Use binocular camera disparity maps to build a point cloud model of the accident scene; The transfer of rotational momentum during cargo dropping is simulated based on the physics engine.

[0012] Specifically, it also includes: The Beidou / GPS dual-mode positioning module of the distribution terminal is called to match the location data with the electronic fence database to identify the scene attributes of accidents occurring on steps, slopes or no-parking areas.

[0013] According to a second aspect of the present invention, there is provided an intelligent evidence collection system for courier delivery accidents, comprising: Embedded sensing modules, including an inertial measurement unit built into the delivery terminal, are used to monitor the courier's movement status in real time; A multi-source evidence collection module, including the delivery terminal's front and rear cameras and microphones, collects multi-angle video streams, depth map data, and ambient audio from the accident scene; The control module is used to automatically activate the multi-source forensics module when a sudden acceleration change exceeding a preset threshold is detected; it is used to perform real-time frame sequence analysis on the video stream and identify the courier's body movements and the trajectory of the cargo being dropped through the spatiotemporal attention mechanism; it is used to detect key event points of collision sounds and human quarrels based on the audio spectrum features collected by the microphone; it is used to cross-modally align IMU data, video motion features and audio event points to construct a causal event graph with synchronized timestamps; it calls a pre-trained accident responsibility quantification model to output the courier's responsibility coefficient, third-party responsibility coefficient and environmental factor weights according to the causal event graph; it is used to encrypt the multi-source forensics data and its timestamp and geographic fence information and write them into the blockchain distributed ledger; and it generates an intelligent forensics report containing a three-dimensional scene reconstruction animation, a responsibility coefficient matrix and a blockchain evidence number.

[0014] According to a third aspect of the present invention, there is provided an electronic device comprising: a memory; and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the above-mentioned intelligent evidence collection method for courier delivery accidents is implemented.

[0015] Beneficial effects: (1) Innovation in evidence capture capabilities: Through IMU threshold linkage and simultaneous activation of multiple sensors, we can break through the timing blind spots of passive evidence collection and ensure dynamic coverage of the entire accident process; (2) Intelligent responsibility determination: The causal event graph constructed by cross-modal alignment eliminates spatiotemporal bias, and the pre-trained responsibility model integrates physical laws and behavioral patterns to output accurate tripartite responsibility coefficients; (3) Strengthening of legal effectiveness: The combination of blockchain evidence storage and three-dimensional physical engine enables video movements, environmental parameters, and voiceprint features to form an unalterable closed-loop evidence chain.

[0016] (4) Embedded perception module achieves millisecond-level response: The six-axis IMU and dual-camera hardware work together to meet the real-time processing requirements of high-dynamic scenes; (5) Strong terminal compatibility: It can adapt to the hardware configuration of mainstream distribution terminals, and Beidou / GPS dual-mode positioning ensures scene attribute recognition in complex urban environments; (6) Excellent power consumption control: The full-function module is activated only when the acceleration changes suddenly, reducing energy consumption by 70% compared with the continuous recording solution.

[0017] Overall improvement effect: It has greatly improved the efficiency of accident responsibility determination, greatly shortened the judicial dispute resolution cycle, and significantly optimized the operating costs of express delivery companies. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of an intelligent evidence collection method for courier delivery accidents provided in a specific embodiment of the present invention; Figure 2 It is a schematic diagram of the system composition of the intelligent evidence collection system for courier delivery accidents provided in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is clearly and completely described below in conjunction with the drawings of the present invention. Based on the embodiments in this application, other similar embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of this application. In addition, the directional words mentioned in the following embodiments, such as "up", "down", "left", "right", etc., are only reference to the directions of the drawings. Therefore, the directional words used are used to illustrate rather than limit the invention.

[0020] The present invention will be further described below with reference to the accompanying drawings and preferred embodiments.

[0021] See also Figure 1 The present invention provides an intelligent evidence collection method for courier delivery accidents, comprising: S100 monitors the courier's motion status in real time through the inertial measurement unit built into the delivery terminal. When a sudden acceleration change exceeding a preset threshold is detected, the multi-source evidence collection module is automatically activated, synchronously triggering the front camera, rear camera and microphone of the delivery terminal to collect multi-angle video streams, depth map data and ambient audio at the accident scene.

[0022] Specifically, step S100 includes: ①IMU activation threshold setting: Use a six-axis IMU (including a three-axis accelerometer and a three-axis gyroscope), with a sampling frequency of ≥100 Hz. Acceleration mutation threshold: Activation is triggered when the three-axis composite acceleration value meets the following conditions: in: , , : The instantaneous measurement value of the accelerometer on the X / Y / Z axis (unit: g) : triaxial composite acceleration modulus; : Acceleration change rate (unit: g / s).

[0023] Optimized value based on: Based on IMU data analysis of 500 real express delivery accidents: Fall scenario: peak acceleration 4.2±1.3g, rate of change 28±7g / s; Throwing scenario: peak acceleration 5.8±2.1g, rate of change 32±9g / s; The threshold is set higher than the upper limit of normal movement (maximum 2.5g / 12g / s for running).

[0024] ② The multi-source data acquisition parameters are shown in Table 1: Table 1 Multi-source data acquisition parameters S200: Perform real-time frame sequence analysis on the video stream, identify the courier's body movements and the trajectory of the cargo dropped through the spatiotemporal attention mechanism; detect key event points of collision sounds and human quarrels based on the audio spectrum features collected by the microphone.

[0025] Specifically, the spatiotemporal attention mechanism includes: Track the key points of the human skeleton in the video frame sequence and calculate the rate of change of the courier's torso inclination; The YOLO-X model is used to detect the trajectory of the parabola and mark the coordinates of the collision area with the ground.

[0026] It should be further explained that, regarding step S200, the solution designed by the present invention includes: ① Design a human skeleton tracking solution: The Open Pose model is used to extract 17 key points and calculate the torso inclination angle: in: : Left / right hip joint key point coordinates (pixels); : shoulder / hip center point coordinates; : Angle between the torso and the vertical (unit: degree).

[0027] Decision threshold: Imbalanced state: Normal walking ; Basis: Biomechanical experiments show When the center of gravity shifts beyond the stability boundary.

[0028] ②Cargo trajectory detection: The YOLO-V10 model has an input size of 640×640 and a confidence threshold of 0.7. Parabolic trajectory equation: in: is the initial velocity of the hand leaving the machine (calculated by the displacement difference of adjacent frames); Represents the real-time position coordinates of the package in three-dimensional space; represents the initial position vector; represents the time variable; Represents the gravitational acceleration vector.

[0029] ③Audio event detection: Key acoustic feature extraction: Specifically: The collision sound judgment condition is >80dB; The judgment conditions for quarreling are .

[0030] Specifically, the physical meanings and numerical setting basis of the components of the above impact sound formula are shown in Table 2: Table 2 Physical meaning and numerical settings of the components of the impact sound formula Specifically, the physical meanings and numerical setting basis of the components of the above-mentioned quarrel sound formula are shown in Table 3: Table 3 Physical meaning and numerical settings of the components of the quarrel sound formula S300: Cross-modally align IMU data, video motion features, and audio event points to construct a causal event graph with synchronized timestamps; call a pre-trained accident responsibility quantification model to output the courier responsibility coefficient, third-party responsibility coefficient, and environmental factor weights based on the causal event graph.

[0031] Specifically, the cross-modal alignment adopts a dynamic time warping algorithm, and the alignment method satisfies: The deviation between the peak moment of IMZ axial acceleration and the frame when the cargo leaves the hand in the video is controlled within ±100ms; The time synchronization error between the audio voiceprint mutation point and the video collision frame is less than 50% of the audio frame length.

[0032] It should be further explained that the multimodal alignment solution designed in the present invention includes: Design of Dynamic Time Warping (DTW) algorithm: Define the distance function: in: : cumulative distance matrix; : Euclidean distance of eigenvector ; : Video feature vector ; : Y coordinate of the bottom edge of the wrapping bounding box; , : horizontal / vertical movement speed of the package; : IMU feature vector ; : Z-axis acceleration; : Y-axis angular velocity; Eigenvector ; Synchronization constraints: .

[0033] Specifically, the training method of the accident responsibility quantification model includes: Use a generative adversarial network to synthesize 100,000 delivery accident scene samples; A weakly supervised learning framework is used to label the ternary responsibility labels of courier error, third-party collision, and slippery road surface.

[0034] It should be further explained that, regarding the accident responsibility quantification model, the scheme designed by the present invention includes: ①GAN sample generation: Generator structure: 5-layer ResNet, latent space dimension z=128; Loss function: in: : Discriminator; : generator; : Gradient penalty coefficient (optimized value 0.001); : Noise distribution (Gaussian distribution N(0,1)).

[0035] 2. Calculation of responsibility coefficient Graph Neural Network Inference: in: is the courier’s responsibility coefficient; is the third party's liability coefficient; is the environmental responsibility coefficient; : Accident feature vector, including: maximum acceleration, falling height, road slope; : first layer weight (256×300); : output layer weight (300×3); Softmax: normalization processing; Penalty coefficient Preventing exploding gradients).

[0036] Specifically, it also includes: When a human quarrel is detected, the voice emotion analysis module is activated to extract the anger index and threat keyword frequency in the voiceprint features, and generate a chain of evidence of provocative behavior by a third party.

[0037] Specifically, the speech emotion analysis adopts: Mel-frequency cepstral coefficients combined with dual-stream feature extraction of BERT speech encoder; Localizing the temporal intervals of abusive sentences via gradient class activation mapping.

[0038] It should be further explained that regarding speech emotion analysis, the solution designed by the present invention includes: 1. Design a dual-stream feature extraction solution: Dual-stream features: MFCC+BERT speech coding; Among them, MFCC is the Mel-frequency cepstral coefficient, which is used to characterize acoustic physical features: fundamental frequency, formant, etc.; BERT speech encoding is a Transformer-based semantic vector used to capture language context information (such as threat semantics).

[0039] 2. Design the anger index model: Anger Index: ; Specifically, the meanings and technical functions of the components of the anger index formula are shown in Table 4; Table 4 Meaning and technical function of the components of the anger index formula Weight setting: (insulting words like "break" / "pay money", etc.), (The tone rises sharply).

[0040] It should be further explained that the weight The allocation logic is shown in Table 5: Table 5 Meaning and technical function of the components of the anger index formula Judgment rules: Common disputes: ; Vicious conflict: (Triggering additional evidence).

[0041] ④ Voice emotion analysis: The following is an example to illustrate: Scenario: The courier and the recipient have an argument; ① Event: Timeline: 0-2.4 seconds (8 300ms segments); ②Key events: t3: The recipient shouts "broken", and we get w3 = 0.6, ‖Grad-CAM‖ = 0.82; t5: The tone rises sharply (Δf = 52 Hz / ms), resulting in w5 = 0.4 and ‖Grad-CAM‖ = 0.73; Other periods: ‖Grad-CAM‖≈0.1 (emotional stability).

[0042] ③Anger index calculation: System response: A value >30 is considered a malicious conflict, automatically triggering video evidence storage; Technical advantages: 1. Semantic-acoustic dual-driver: MFCC captures intonation changes + BERT understands offensive semantics; 2. Strengthening of legal effectiveness: The weight setting matches the criteria for determining "verbal insult" in the Civil Code; 3. Resource optimization: Only data from high-weight time periods is encrypted and stored (blockchain storage capacity is greatly reduced).

[0043] S400: Encrypt the multi-source forensic data and its timestamp and geo-fence information and write them into the blockchain distributed ledger; generate an intelligent forensic report including a three-dimensional scene reconstruction animation, a responsibility coefficient matrix, and a blockchain evidence storage number.

[0044] Specifically, the generation of the 3D scene reconstruction animation includes: Use binocular camera disparity maps to build a point cloud model of the accident scene; The transfer of rotational momentum during cargo dropping is simulated based on the physics engine.

[0045] It should be further explained that, regarding the reconstruction of the three-dimensional scene, the solution designed by the present invention includes designing a collision physics model: in: : total energy loss (unit: joule); : Material elastic coefficient - carton: N / m (GB / T6544 standard test); : shape (measured by depth map); : Friction coefficient - asphalt pavement: (GB50009-2012); : Package quality (obtained from the waybill database); : acceleration due to gravity; : sliding distance (by visual ranging); : Rotational energy conversion coefficient (optimized value 0.4); : Package moment of inertia - rectangular parallelepiped: in: represents the length of the cuboid, h represents the height of the cuboid, and m represents the mass of the cuboid; : Angular velocity (calculated by differencing consecutive frames).

[0046] Specifically, it also includes: The Beidou / GPS dual-mode positioning module of the distribution terminal is called to match the location data with the electronic fence database to identify the scene attributes of accidents occurring on steps, slopes or no-parking areas.

[0047] Furthermore, the method further comprises: 1. Dynamic correction of responsibility coefficient: Input: scene labels matched by geofence database; The correction rules are shown in Table 6: Table 6 Amendment Rules Output: Updated responsibility coefficient matrix .

[0048] The following is an example to illustrate: Slope gradient 12°, Increased by 0.6 (original value 0.24, revised to 0.84).

[0049] 2. Enhanced 3D physical modeling ①Stairs scene: Simulate the multi-level collision trajectory of a falling package: in: The total energy loss is calculated to quantify the extent of collision damage. : Number of step levels (identified by depth map), the number of steps is identified by ToF depth map; : The height of the i-th drop is calculated by the binocular camera parallax; : step inclination, measured by the IMU gyroscope; is the elastic coefficient of the packaging material, the default value (corrugated box: k = 1200N / m); The quality of the package is obtained from the waybill database; g is the acceleration due to gravity; is the friction coefficient of the step surface, which is preset according to the material (cement step: μ=0.3); is the sliding distance of the i-th step, calculated by tracking the displacement of the wrapping bounding box.

[0050] ②Slope scene: Calculation of sliding distance: : Dynamic friction coefficient (wet road μ=0.15).

[0051] in: is the total sliding distance of the package on the slope. The calculation result is used to locate the collision point and determine the responsibility; To include the initial velocity when sliding, it is calculated by the displacement difference of adjacent video frames (accuracy ±0.2m / s); g is the acceleration due to gravity; μ is the dynamic friction coefficient, which is set to 0.15 for wet roads, based on the road friction coefficient standard "JT / T715-2008" (the measured value for wet asphalt is 0.12-0.18); θ is the angle between the slope and the horizontal plane, measured by the IMU gyroscope; is the cosine of the slope angle, which is used to calculate the vertical component effect of friction; is the sine of the slope angle, which is used to calculate the effect of the glide component of gravity.

[0052] ③No parking area: Superimpose illegal stay duration parameters: , triggering a full responsibility judgment.

[0053] 3. Blockchain-based targeted evidence storage.

[0054] ①Key evidence storage in prohibited parking areas: Store the electronic map geo-fence registration number + courier stay timestamp; ②Slope / stairs scene: Fixed key evidence: hash value of measurement certificate for slope / step height (compliant with GB50026-2020 engineering measurement standard).

[0055] 4. Intelligent report generation: ① Scene identification enhancement: Annotate the 3D reconstruction animation with a geographic risk label (for example, a slope with a gradient of 12°).

[0056] ② Rearrange the responsibility matrix, as shown in Table 7: Table 7 Schematic diagram of responsibility matrix rearrangement The report concludes: "The accident occurred on a slope (slope 12°), environmental responsibility accounts for 84%, and it is recommended that the courier be exempted from compensation."

[0057] Technical solution advantages: 1. Dynamic responsibility mapping: Scenario attributes directly drive responsibility weights (step height / slope, linear increments); No-parking zones trigger penalty weights (breaking the Softmax constraint).

[0058] 2. Physics engine enhances credibility: The slope sliding formula is based on Newtonian mechanics, which greatly reduces calculation errors; The step collision energy accumulation model was verified through ABAQUS simulation.

[0059] 3. Closed loop of legal effectiveness: National standards for evidence storage; 3D animation + correction coefficient matrix constitutes a strong chain of evidence.

[0060] See also Figure 2 The present invention provides another embodiment, which provides an intelligent evidence collection system for courier delivery accidents. The intelligent evidence collection system for courier delivery accidents includes: The embedded sensing module 100 includes an inertial measurement unit built into the delivery terminal, which is used to monitor the courier's movement status in real time; The multi-source evidence collection module 200 includes a front-facing camera, a rear-facing camera, and a microphone at the delivery terminal, which respectively collect multi-angle video streams, depth map data, and ambient audio at the accident scene; The control module 300 is used to automatically activate the multi-source forensics module when a sudden acceleration change exceeding a preset threshold is detected; to perform real-time frame sequence analysis on the video stream, and identify the courier's body movements and the trajectory of the cargo being dropped through a spatiotemporal attention mechanism; to detect key event points of collision sounds and human voice quarrels based on the audio spectrum features collected by the microphone; to perform cross-modal alignment of IMU data, video motion features, and audio event points to construct a causal event graph with synchronized timestamps; to call a pre-trained accident responsibility quantification model to output the courier's responsibility coefficient, third-party responsibility coefficient, and environmental factor weights based on the causal event graph; to encrypt the multi-source forensics data and its timestamp and geographic fence information and write them into the blockchain distributed ledger; and to generate an intelligent forensics report containing a three-dimensional scene reconstruction animation, a responsibility coefficient matrix, and a blockchain evidence storage number.

[0061] In a preferred embodiment, the present application further provides an electronic device, comprising: A memory; and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the method for intelligent evidence collection of courier delivery accidents is implemented. The computer device can be broadly defined as a server, a terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, a memory, a network interface, a communication interface, etc. connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing and / or control capabilities. The memory of the computer device may include a non-volatile storage medium and an internal memory. An operating system, a computer program, etc. may be stored in or on the non-volatile storage medium. The internal memory can provide an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface and the communication interface of the computer device can be used to connect and communicate with external devices via a network. When the computer program is executed by the processor, the steps of the method of the present invention are performed.

[0062] The present invention can be implemented as a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the method of an embodiment of the present invention to be performed. In one embodiment, the computer program is distributed on a plurality of computer devices or processors coupled to a network so that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, can be performed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations can be performed by one or more computer devices or processors, and one or more other method steps / operations can be performed by one or more other computer devices or processors. One or more computer devices or processors can perform a single method step / operation, or perform two or more method steps / operations.

[0063] Those skilled in the art will appreciate that the method steps of the present invention can be performed by instructing related hardware, such as a computer device or processor, through a computer program. The computer program can be stored in a non-transitory computer-readable storage medium, and when the computer program is executed, the steps of the present invention are performed. Depending on the circumstances, any reference herein to memory, storage, database, or other media may include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state disk, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0064] The various technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification as long as such combination does not conflict.

[0065] The specific embodiments of the present invention described above do not limit the scope of protection of the present invention. Any other corresponding changes and modifications made based on the technical concept of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. An intelligent evidence collection method for courier delivery accidents, characterized by: The method comprises: S100 monitors the courier's motion status in real time through the delivery terminal's built-in inertial measurement unit. When a sudden acceleration exceeding a preset threshold is detected, the multi-source evidence collection module is automatically activated, synchronously triggering the delivery terminal's front and rear cameras and microphone to collect multi-angle video streams, depth map data, and ambient audio from the accident scene. S200: Perform real-time frame sequence analysis on the video stream, identify the courier's body movements and the trajectory of the cargo drop through the spatiotemporal attention mechanism; detect key event points of collision sounds and human voice arguments based on the audio spectrum characteristics collected by the microphone; S300, cross-modally aligning IMU data, video motion features, and audio event points to construct a causal event graph with synchronized timestamps; calling a pre-trained accident responsibility quantification model to output the courier responsibility coefficient, third-party responsibility coefficient, and environmental factor weights based on the causal event graph; S400: Encrypt the multi-source forensic data and its timestamp and geo-fence information and write them into the blockchain distributed ledger; generate an intelligent forensic report including a three-dimensional scene reconstruction animation, a responsibility coefficient matrix, and a blockchain evidence storage number.

2. The intelligent evidence collection method for courier delivery accidents according to claim 1 is characterized in that: The spatiotemporal attention mechanism specifically includes: Track the key points of the human skeleton in the video frame sequence and calculate the rate of change of the courier's torso inclination; The YOLO model is used to detect the trajectory of the parabola and mark the coordinates of the collision area with the ground.

3. The intelligent evidence collection method for courier delivery accidents according to claim 1 is characterized in that: The cross-modal alignment adopts the dynamic time warping algorithm, and the alignment method satisfies: The deviation between the peak moment of IMZ axial acceleration and the frame when the cargo leaves the hand in the video is controlled within ±100ms; The time synchronization error between the audio voiceprint mutation point and the video collision frame is less than 50% of the audio frame length.

4. The intelligent evidence collection method for courier delivery accidents according to claim 1 is characterized in that: The training method of the accident responsibility quantification model includes: Use a generative adversarial network to synthesize 100,000 delivery accident scene samples; A weakly supervised learning framework is used to label the ternary responsibility labels of courier error, third-party collision, and slippery road surface.

5. The intelligent evidence collection method for courier delivery accidents according to claim 1 is characterized in that: Also includes: When a human quarrel is detected, the voice emotion analysis module is activated to extract the anger index and threat keyword frequency in the voiceprint features, and generate a chain of evidence of provocative behavior by a third party.

6. The intelligent evidence collection method for courier delivery accidents according to claim 5 is characterized in that: The speech emotion analysis adopts: Mel-frequency cepstral coefficients combined with dual-stream feature extraction of BERT speech encoder; Localizing the temporal intervals of abusive sentences via gradient class activation mapping.

7. The intelligent evidence collection method for courier delivery accidents according to claim 1 is characterized in that: The generation of the three-dimensional scene reconstruction animation includes: Use binocular camera disparity maps to build a point cloud model of the accident scene; The transfer of rotational momentum during cargo dropping is simulated based on the physics engine.

8. The intelligent evidence collection method for courier delivery accidents according to claim 1 is characterized in that: Also includes: The Beidou / GPS dual-mode positioning module of the distribution terminal is called to match the location data with the electronic fence database to identify the scene attributes of accidents occurring on steps, slopes or no-parking areas.

9. An intelligent evidence collection system for courier delivery accidents, characterized in that: include: Embedded sensing modules, including an inertial measurement unit built into the delivery terminal, are used to monitor the courier's movement status in real time; A multi-source evidence collection module, including the delivery terminal's front and rear cameras and microphones, collects multi-angle video streams, depth map data, and ambient audio from the accident scene; The control module is used to automatically activate the multi-source forensics module when it detects a sudden acceleration change exceeding a preset threshold. It is used to perform real-time frame sequence analysis on the video stream, identifying the courier's body movements and the trajectory of the dropped goods through a spatiotemporal attention mechanism. Based on the audio spectrum characteristics collected by the microphone, it detects key event points such as collision sounds and human voice arguments. It is used to perform cross-modal alignment of IMU data, video motion features, and audio event points to construct a causal event graph with synchronized timestamps; it calls a pre-trained accident responsibility quantification model to output the courier responsibility coefficient, third-party responsibility coefficient, and environmental factor weights based on the causal event graph; Used to encrypt multi-source forensic data and its timestamps and geo-fence information and write them to the blockchain distributed ledger; Generate an intelligent forensic report including a 3D scene reconstruction animation, a responsibility coefficient matrix, and a blockchain evidence number.

10. An electronic device, characterized in that: include: Memory; and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the intelligent evidence collection method for courier delivery accidents according to any one of claims 1 to 8 is implemented.

Images