Evidence obtaining method and system based on multi-camera association verification, vehicle and aircraft

Abstract

The invention discloses an evidence obtaining method and system based on multi-camera association verification, a vehicle and an aircraft, and the method comprises the steps: carrying out the continuous shooting of a target object or a target position through a first camera when a vehicle-mounted or airborne evidence obtaining device is used for obtaining evidence, and carrying out the continuous shooting of the internal or external environment of a carrier through a second camera; the video shot by the first camera and the local inertial measurement data and positioning data of the first camera are used as a source data set of an evidence obtaining video, and the video shot by the second camera and the local inertial measurement data and positioning data of the second camera are used as a source data set of a verification video; the source data set of the evidence obtaining video and the source data set of the verification video are jointly used for executing a preset cross verification mechanism when the authenticity effectiveness of the video shot by the first camera needs to be verified. According to the invention, the credibility and reliability of the electronic evidence can be enhanced.

Description

Technical Field

[0001] This application relates to the field of forensics technology, and in particular to a mobile forensics method, system, vehicle, and aircraft based on multi-camera correlation verification. Background Technology

[0002] Currently, electronic evidence collection technology has gained widespread recognition and application in judicial evidence collection, providing a more scientific and convenient means of evidence gathering. However, in specific judicial evidence collection practices, the application of electronic evidence collection technology still faces many problems, and new problems continue to emerge.

[0003] Specifically, traditional evidence collection methods usually rely on a single and limited data source, which limits their perspective and makes it difficult to capture the details of the evidence target and the surrounding environment at the same time. This results in a gap in the "process integrity" of the evidence chain. Furthermore, the credibility of traditional electronic evidence is highly dependent on external and lagging certification by public authorities (such as notarized certificates issued by notary offices). It is impossible to prove that there was no tampering, interruption or human intervention during the recording period, making it difficult for the obtained electronic evidence to form a legally valid evidence chain. Summary of the Invention

[0004] In view of this, embodiments of this application provide a mobile evidence collection method, system, vehicle, and aircraft based on multi-camera association verification, which can improve the integrity of the evidence collection process and perform effective internal verification of the evidence collection video.

[0005] In a first aspect, embodiments of this application provide a mobile forensics method based on multi-camera correlation verification, implemented using vehicle-mounted or airborne forensics equipment, which includes at least a first camera and a second camera. The method includes: when using the vehicle-mounted or airborne forensics equipment, continuously filming the target object or target location using the first camera, and continuously filming the internal or external environment of the carrier using the second camera; using the video captured by the first camera along with the local inertial measurement data and positioning data of the first camera as the source dataset of the forensics video, and using the video captured by the second camera along with the local inertial measurement data and positioning data of the second camera as the source dataset of the verification video; the source dataset of the forensics video and the verification... The source datasets of the videos are used together to execute a preset cross-validation mechanism when it is necessary to verify the authenticity of the video captured by the first camera. The cross-validation mechanism includes: identifying at least one set of corresponding physical feature data in the source datasets of the evidence-collecting video and the source datasets of the verification video; reconstructing the physical feature model of the environment in which the first camera is located during the first time period of the evidence-collecting process based on the at least one set of corresponding physical feature data; reconstructing the physical feature model of the environment in which the second camera is located during the first time period of the evidence-collecting process based on the at least one set of corresponding physical feature data; if the two reconstructed physical feature models meet the requirement of mutual verification, then the cross-validation of the set of physical feature data for the first time period is successful.

[0006] Secondly, embodiments of this application provide a multi-camera mobile evidence collection system, which includes at least two cameras and a movable carrier. The multi-camera mobile evidence collection system is used to implement the steps of the mobile evidence collection method based on multi-camera association verification as described in the first aspect.

[0007] Thirdly, embodiments of this application provide a vehicle with at least two cameras installed inside and outside the vehicle, the vehicle being used to implement the steps of the mobile forensics method based on multi-camera association verification as described in the first aspect.

[0008] Fourthly, embodiments of this application provide an aircraft equipped with at least two cameras installed inside and outside the aircraft, which is used to implement the steps of the mobile forensics method based on multi-camera association verification as described in the first aspect.

[0009] Fifthly, embodiments of this application provide an electronic device, including: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the steps of the mobile forensics method based on multi-camera association verification as described in the first aspect.

[0010] In a sixth aspect, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps of the mobile forensics method based on multi-camera association verification as described in the first aspect.

[0011] In a seventh aspect, embodiments of this application provide a computer program product, which is stored in a non-volatile storage medium, and when executed by a processor, the computer program product implements the steps of the mobile forensics method based on multi-camera association verification as described in the first aspect.

[0012] Eighthly, embodiments of this application provide a chip including a processor and a communication interface, the communication interface and the processor being coupled, the processor being used to run programs or instructions to implement the steps of the mobile forensics method based on multi-camera association verification as described in the first aspect.

[0013] This application provides a mobile evidence collection method, system, vehicle, and aircraft based on multi-camera correlation verification. It utilizes a first camera and a second camera to simultaneously acquire video and physical data such as inertial measurement and positioning. Specifically, the first camera captures details of the evidence target, while the second camera simultaneously records the internal or external environmental state of the carrier. This compensates for the shortcomings of traditional evidence collection in terms of "process integrity," enhancing the integrity and relevance of evidence and addressing the fundamental deficiency in proving process integrity in traditional single-perspective methods. Based on this, a cross-verification mechanism based on physical feature data model reconstruction can reconstruct the physical feature model of the environment in which the cameras are located using two sets of data and verify their mutual corroboration. The credibility authentication of evidence shifts from traditional "external institution verification" to "real-time self-verification by technical means." Its credibility no longer solely relies on external, lagging authentication methods, effectively improving the strength and efficiency of proving the original authenticity and process integrity of electronic evidence, thereby enhancing the tamper resistance of the evidence. Furthermore, by relying on vehicle-mounted / airborne mobile platforms and their local sensor data, this solution enhances environmental adaptability, operational flexibility, and execution reliability in complex, mobile evidence collection scenarios, enabling mobile evidence collection activities to have broader applicability while ensuring the legal validity of evidence. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings of the embodiments of this application will be briefly described below.

[0015] Figure 1 This is a flowchart illustrating a mobile forensics method based on multi-camera association verification provided in an embodiment of this application. Figure 2 This is an exemplary schematic diagram of a movable carrier provided in an embodiment of this application; Figure 3This is an exemplary structural diagram of a multi-camera mobile evidence collection system provided in an embodiment of this application; Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0016] The principles and spirit of this application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided to make the principles and spirit of this application clearer and more thorough, enabling those skilled in the art to better understand and implement the principles and spirit of this application. The exemplary embodiments provided herein are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments described herein without inventive effort are within the scope of protection of this application.

[0017] In this document, terms such as first, second, and third are used only to distinguish one entity (or operation) from another, and are not intended to require or imply any order or relationship between these entities (or operations).

[0018] The following is a brief description of the concepts and technical terms that may be involved in the embodiments of this application.

[0019] A Trusted Timestamp (TSA) is an electronic certificate issued by an authoritative organization in accordance with international standards to prove the existence, integrity, and immutability of electronic data.

[0020] Before describing the technical solutions provided in the embodiments of this application, in order to facilitate understanding of the embodiments of this application, this application first specifically explains the problems existing in the related technologies: Traditional electronic data forensics often lacks effective real-time verification mechanisms, making it possible for critical evidence such as videos to be tampered with before TSA intervention, thus affecting the fairness of the entire case and the judgment. Specifically, traditional electronic data forensics often relies on recording with a single device, which has a limited perspective and cannot simultaneously capture the details of the evidence target and the behavior of the evidence-gathering subject. This results in a gap in the "process integrity" of the evidence chain, making it difficult to comprehensively and in real-time record various information during the evidence gathering process, and failing to form an effective evidence verification system, thus limiting the methods of evidence gathering. Furthermore, the credibility of traditional electronic evidence is highly dependent on external, delayed certification by public authorities (such as notarized certificates issued by notary offices), and cannot prove that there was no tampering, interruption, or human intervention during the recording.

[0021] In view of the inventors’ above-mentioned research findings, this application provides a mobile forensics method, system, vehicle and aircraft based on multi-camera correlation verification, which aims to solve the problem of low authenticity and reliability of evidence in traditional electronic data forensics.

[0022] The mobile forensics method based on multi-camera association verification provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0023] Figure 1 This is a flowchart illustrating a mobile forensics method based on multi-camera correlation verification according to an embodiment of this application. The executing entity of this mobile forensics method based on multi-camera correlation verification can be a multi-camera mobile forensics system (hereinafter referred to as "forensics system"). Specifically, this method can be implemented using vehicle-mounted or airborne forensics equipment in the forensics system. The forensics equipment includes at least a first camera and a second camera. It should be noted that the above-mentioned executing entity and application scenario do not constitute a limitation on this application.

[0024] like Figure 1 As shown, the mobile forensics method based on multi-camera association verification provided in this application embodiment may include steps 110-130.

[0025] Step 110: When collecting evidence using vehicle-mounted or airborne evidence collection equipment, continuously photograph the target object or target location using the first camera, and continuously photograph the internal or external environment of the carrier using the second camera. Step 120: Use the video captured by the first camera, along with the local inertial measurement data and positioning data of the first camera, as the source dataset for the evidence video; use the video captured by the second camera, along with the local inertial measurement data and positioning data of the second camera, as the source dataset for the verification video. Step 130: Based on the source dataset of the evidence video and the source dataset of the verification video, execute a preset cross-verification mechanism when it is necessary to verify the authenticity and validity of the video captured by the first camera.

[0026] The cross-validation mechanism includes: Identify at least one set of corresponding physical feature data existing in the source dataset of the evidence-collecting video and the source dataset of the verification video; based on at least one set of corresponding physical feature data, reconstruct the physical feature model of the environment where the first camera is located during the first time period of the evidence-collecting process; based on at least one set of corresponding physical feature data, reconstruct the physical feature model of the environment where the second camera is located during the first time period of the evidence-collecting process; if the two reconstructed physical feature models meet the requirement of mutual verification, then the cross-validation of the set of physical feature data for the first time period is successful.

[0027] The mobile evidence collection method based on multi-camera correlation verification provided in this application utilizes a first camera and a second camera to simultaneously collect video and physical data such as inertial measurement and positioning. Specifically, the first camera captures details of the evidence target, while the second camera simultaneously records the internal or external environmental state of the carrier, compensating for the shortcomings of traditional evidence collection in terms of "process integrity," enhancing the integrity and relevance of evidence, and addressing the fundamental deficiency of traditional single-perspective proof of process integrity. Based on this, a cross-verification mechanism based on physical feature data model reconstruction can reconstruct the physical feature model of the environment where the cameras are located using two sets of data and verify their mutual corroboration. The credibility authentication of evidence shifts from traditional "external institution verification" to "real-time self-verification by technical means." Its credibility no longer solely relies on external, lagging authentication methods, effectively improving the strength and efficiency of proving the original authenticity and process integrity of electronic evidence, thereby enhancing the tamper resistance of evidence. Furthermore, this solution relies on vehicle / airborne mobile platforms and their local sensor data, improving environmental adaptability, operational flexibility, and execution reliability in complex, mobile evidence collection scenarios, enabling mobile evidence collection activities to have broader applicability while ensuring the legal validity of evidence.

[0028] The specific implementation of the above steps will be described in detail below with reference to specific embodiments.

[0029] In step 110, the first camera (main evidence-gathering camera) continuously captures images of the designated target object or location to generate evidence-gathering video. This target object is the evidence-gathering object, such as an accident scene or physical evidence. The second camera (auxiliary verification camera) simultaneously captures images of the internal or external environment of the carrier to generate verification video. This carrier is specifically a movable carrier, such as a vehicle, aircraft, or other evidence-gathering equipment. The first and second cameras have different fields of view; this dual-view design ensures that both target details and the state of the evidence-gathering environment are recorded, constructing a more comprehensive on-site record.

[0030] According to embodiments of this application, optionally, vehicle-mounted or airborne evidence collection equipment includes evidence collection equipment mounted on any of manned vehicles, unmanned vehicles, manned aircraft, drones, robots, or embodied intelligent agents.

[0031] According to embodiments of this application, optionally, the camera includes at least one of the following: a monitoring device for the interior space of a vehicle, a monitoring device for people inside the vehicle, a monitoring device for the exterior space of a vehicle, a monitoring device for people outside the vehicle, a dashcam, an airborne monitoring device, or a monitoring device mounted on an embodied intelligent agent.

[0032] Thus, by clearly defining the diverse types of carriers and cameras, this solution can adapt to various complex evidence collection scenarios, from ground to air, from fixed routes to autonomous mobility, thereby improving the system's practicality and deployment flexibility.

[0033] In step 120, each camera's local sensor unit synchronously records inertial measurement unit (IMU, including acceleration, angular velocity, etc.), positioning data (such as GPS and BeiDou coordinates), and high-precision clock data. These data correspond strictly to the video frames and together constitute the source dataset.

[0034] Optionally, the verification process based on the source dataset of the evidence-collecting video and the source dataset of the verification video can be performed by the evidence-collecting device. Alternatively, the multi-camera mobile evidence-collecting system includes a server with a trusted authentication application, whereby the evidence-collecting device can upload the collected source dataset to the server, and the server can perform the aforementioned verification process.

[0035] Optionally, the source dataset collected in this application can be sent to the server in two ways. Method one is real-time transmission: video and sensor data are streamed to the server in real-time via a wireless network, suitable for scenarios with good network conditions and requiring immediate verification. Method two is asynchronous secure transmission: in scenarios with poor network conditions or where data security is emphasized, the source dataset can first be cached and integrated in a Trusted Execution Environment (TEE) or security sandbox locally on the forensic device or removable carrier. Then, the data is encrypted using encryption algorithms (such as RSA or AES), and a digital signature is generated (e.g., calculating a checksum and signing based on a hash algorithm such as SHA-256), forming an immutable signed data packet before being asynchronously sent to the server. This method effectively prevents data from being stolen or tampered with during transmission.

[0036] In step 130, the source dataset of the evidence-collecting video and the source dataset of the verification video are used together to execute a preset cross-validation mechanism when it is necessary to verify the authenticity of the video captured by the first camera. The evidence-collecting system extracts one or more sets of corresponding physical feature data from the source datasets of the evidence-collecting video and the verification video, respectively. The physical feature data may include at least one of spatial relationship data, illumination relationship data, audio signal data, motion feature data, and scene feature data.

[0037] The core objective of cross-validation is to verify whether physical feature models reconstructed from data collected from different perspectives (the first camera and the second camera) within the same time period depict the same consistent physical environment. Its basic principle is that if the two video streams and their associated sensor data are authentic, continuous, and untampered with, then the environmental model reconstructed from the physical features extracted from each should satisfy specific spatial, geometric, or physical constraints. Therefore, this "mutual verification" provides a basis for verifying the authenticity of the evidence videos.

[0038] Specifically, the system compares the two reconstructed physical feature models. The verification logic focuses on the consistency constraints between the models, for example: Motion consistency: If the first camera video shows the vehicle accelerating, the second camera's image of the equipment carrier or the person collecting evidence should show a corresponding inertial motion trend (such as leaning backward). The reconstructed model should reflect this motion correlation.

[0039] Spatial consistency: The geometry of the same static scene reconstructed from different perspectives should be able to be aligned after coordinate transformation.

[0040] Optical consistency: Changes in ambient light should be reflected synchronously or in accordance with physical laws in both camera models.

[0041] If the reconstructed models are within the preset tolerance range, such as the feature point matching rate exceeding the threshold or the motion trajectory correlation coefficient being greater than the set value, then they are confirmed to meet the requirements of mutual verification and can pass the above consistency check, and the cross-validation is successful; otherwise, it indicates that the data in this time period may be at risk of being tampered with.

[0042] For example, in the evidence collection for a traffic accident, a cross-verification mechanism can simultaneously utilize multiple physical feature data. Regarding scene feature data, it compares whether the vehicle's body color, model markings, etc., are consistent between the two video streams. Regarding spatial relationship data, it uses Structure for Motion Restoration (SfM) technology to reconstruct a 3D scene model of the accident site based on the videos from both cameras, verifying whether the relative positions of the vehicles involved and road markings match. Regarding lighting relationship data, it analyzes whether the color temperature and intensity change curves of ambient light in the two video streams are synchronized. Regarding audio signal data, it verifies whether the time difference between the appearance of braking sounds and collision sounds in the two audio streams matches the sound wave transmission delay caused by the geographical difference between the two cameras. Thus, this multi-dimensional data cross-verification greatly enhances the credibility of the evidence.

[0043] According to an embodiment of this application, optionally, the physical feature data includes spatial relationship data, and the cross-validation mechanism specifically includes: extracting the three-dimensional coordinates of feature points and the relative spatial positions between objects in the environment captured by the first camera based on the video in the source dataset of the evidence video, to obtain first spatial relationship data, and constructing a first spatial relationship model based on the first spatial relationship data; extracting the three-dimensional coordinates of feature points and the relative spatial positions between objects in the environment where the second camera is located based on the video in the source dataset of the verification video, to obtain second spatial relationship data, and constructing a second spatial relationship model based on the second spatial relationship data; converting the first spatial relationship model and the second spatial relationship model to the same coordinate system, and verifying whether their spatial geometric features meet the preset consistency constraints; if the consistency of the spatial geometric features is higher than the preset spatial correlation threshold, then the cross-validation based on the spatial relationship data is determined to be successful.

[0044] Specifically, spatial relationship data is used to characterize the three-dimensional geometric positions and relative spatial relationships of objects in an evidence-gathering scene, such as distance, angle, and topological relationships (adjacency, containment, etc.). The implementation process is as follows: The server first extracts the three-dimensional coordinates of feature points (such as corner points and edge points) and the relative spatial positions between objects (such as the orientation and distance of object A relative to object B) from the same time period of video from the first camera (evidence-gathering video) and the second camera (verification video). For example, computer vision techniques such as Structure for Motion Restoration (SfM) or Visual Inertial Odometry (VIO) can be used for extraction, and the output is the spatial relationship data. Based on the extracted spatial relationship data, the server constructs a first spatial relationship model and a second spatial relationship model, respectively. These spatial relationship models are used to describe the sparse three-dimensional point cloud or feature frame of the environmental geometry from the perspective of their respective cameras. Subsequently, through coordinate transformation, such as using the relative pose between the two cameras acquired in advance or in real time, the two spatial relationship models are unified to the same coordinate system (such as the world coordinate system). In the verification stage, the three-dimensional positional deviation of corresponding feature points in the two models is calculated, or the consistency of spatial geometric features such as the geometric dimensions and relative distances of specific objects reconstructed from the feature points is compared. If the calculated similarity index (such as the root mean square error (RMSE) of point cloud registration is higher than the preset spatial correlation threshold, it indicates that the spatial relationship models from different perspectives corroborate each other and the cross-validation is successful.

[0045] For example, such as Figure 2As shown, the mobile carrier is a patrol car, collecting evidence at an intersection. The first camera is a dashcam 201, used to record traffic conditions at the intersection, and the second camera is an in-vehicle camera 202 facing the driver, indirectly recording changes in ambient light and shadow outside the vehicle. During the cross-validation phase, the server selects a 5-second time period from when the patrol car brakes to a complete stop. Based on the video from the first camera, the 3D coordinates of feature points such as traffic lights and pedestrian crossings are extracted using SfM technology to construct a first spatial relationship model. Simultaneously, based on the video from the second camera, feature points of external objects as seen from the driver's perspective (such as the approximate location of traffic lights seen through the windshield) are extracted to construct a second spatial relationship model. The two models are transformed to a coordinate system with the front of the car as the origin, and the relative azimuth angles of the traffic lights in the models are compared. If the deviation of the traffic light azimuth angles calculated by the two models is less than a preset threshold (e.g., 3 degrees), the cross-validation based on the spatial relationship data is successful, indicating that the two videos are consistent in spatial geometry.

[0046] In this embodiment, by independently reconstructing and comparing the environmental spatial relationship model under dual perspectives, the geometric consistency of the framing environment can be effectively verified. Tampering is difficult to achieve without vulnerabilities under the spatial geometric constraints of dual perspectives. Therefore, this verification method can effectively detect tampering behaviors such as video splicing and virtual object implantation, thereby enhancing the authenticity and credibility of electronic evidence in the spatial dimension.

[0047] According to an embodiment of this application, optionally, the physical feature data includes audio signal data, and the cross-validation mechanism specifically includes: constructing a first audio feature model reflecting the acoustic environment of the first camera based on the audio signal data contained in the source dataset of the evidence video; constructing a second audio feature model reflecting the acoustic environment of the second camera based on the audio signal data contained in the source dataset of the verification video; verifying whether the first audio feature model and the second audio feature model corroborate each other by determining the cross-correlation between the first audio feature model and the second audio feature model; if the cross-correlation is greater than a preset audio correlation threshold, then the cross-validation based on the audio signal data is determined to be successful.

[0048] Specifically, audio signal data is used to characterize the acoustic features of the evidence-gathering environment. The process is as follows: features reflecting the acoustic environment are extracted from the audio data accompanying the videos from the first and second cameras, such as the sound signature of specific events (e.g., the characteristic frequencies of sirens and brake sounds), the spectral distribution of ambient noise, and the time difference of arrival of audio signals. Based on these features, a first audio feature model and a second audio feature model are constructed to quantify the acoustic environment of their respective microphones. During verification, the cross-correlation between the two audio feature models at the same time interval is calculated. For example, the degree of matching of the acoustic features of the same event (e.g., a siren) in the two audio streams on the time axis is compared, or the spectral similarity of ambient background noise is analyzed. If the calculated cross-correlation coefficient is greater than a preset audio correlation threshold, it indicates that the two audio streams originate from the same acoustic environment, and the cross-verification is successful.

[0049] For example, let's continue with the example of a patrol car collecting evidence at an intersection. During the evidence collection process, a collision occurs at the intersection, producing a loud bang. Features such as sound pressure level and dominant frequency of the collision sound recorded by the first camera (outside the vehicle) are extracted to construct a first audio feature model. Simultaneously, features of the collision sound transmitted through the vehicle's structure and attenuated, recorded by the second camera (inside the vehicle), are extracted to construct a second audio feature model. The cross-correlation between the two models at the moment of the collision is calculated. If the acoustic features are found to be highly synchronized in time and their spectral change trends are consistent (correlation coefficient > 0.8), the verification is successful, indicating that the same acoustic event was recorded both inside and outside the vehicle.

[0050] In this embodiment, the propagation characteristics of audio signals in physical space are utilized to provide an independent acoustic verification dimension for the authenticity of video evidence. Forged audio is difficult to maintain accurate acoustic feature synchronization and correlation under dual-microphone recording. Therefore, this audio verification method can effectively identify tampering techniques such as audio replacement or post-dubbing, thereby enhancing the integrity and reliability of the evidence chain.

[0051] According to embodiments of this application, optionally, the physical feature data includes scene feature data, and the static visual features include key points, edge or texture feature points. The cross-validation mechanism specifically includes: extracting static visual features of the shooting scene corresponding to the first camera based on the video in the source dataset of the evidence video to obtain first scene feature data, and constructing a first scene feature model based on the first scene feature data; extracting static visual features of the shooting scene corresponding to the second camera based on the video in the source dataset of the verification video to obtain second scene feature data, and constructing a second scene feature model based on the second scene feature data; determining the matching degree of corresponding feature points in the first scene feature model and the second scene feature model; if the matching degree is higher than a preset matching degree threshold, then the cross-validation based on the scene feature data is determined to be successful.

[0052] Specifically, scene feature data refers to static visual features extracted from videos, such as keypoints, edge information, and texture feature points, used to describe the appearance of a scene. Keypoints may include, but are not limited to, Scale Invariant Feature Transform (SIFT), Speed-Up Robust Feature Transform (SURF), Directed Fast and Rotated BRIEF (ORB) feature points. The implementation process is as follows: The server extracts static visual features from videos from the first and second cameras respectively. Based on the extracted features, the server constructs a first scene feature model and a second scene feature model, which can be a set of feature descriptors. During the validation phase, a feature matching algorithm (such as nearest neighbor search) is used to calculate the matching degree of corresponding feature points in the two scene feature models, i.e., the proportion of successfully matched feature point pairs to the total number of feature points. If this matching degree is higher than a preset matching degree threshold, the cross-validation based on the scene feature data is considered successful.

[0053] For example, in a scenario involving evidence collection from the facade of a building, a first camera is directly facing the building, while a second camera may capture a portion of the facade from the side. The server extracts features such as window corners and wall textures from the first camera video to construct a first scene feature model. Similar features from overlapping areas are extracted from the second camera video to construct a second scene feature model. After feature matching, if a sufficient number of feature points (e.g., over 70%) in the two models match correctly, it indicates that the two videos captured the same scene, and the scene feature cross-validation is successful.

[0054] In this embodiment, by comparing the static visual features of the scene, the authenticity of the video content in terms of appearance consistency can be effectively verified. Forged content is difficult to perfectly match the real scene in terms of detail texture and feature points. Therefore, this verification method is effective in detecting attacks such as scene replacement and partial image tampering, thus ensuring the authenticity of the video evidence.

[0055] According to an embodiment of this application, optionally, the physical feature data includes illumination relationship data, and the cross-validation mechanism specifically includes: based on the video in the source dataset of the evidence video, analyzing the light source direction, intensity, and shadow change patterns of the environment where the first camera is located to obtain first illumination relationship data, and constructing a first illumination change model based on this; based on the video in the source dataset of the verification video, analyzing the light source direction, intensity, and shadow change patterns of the environment where the second camera is located to obtain second illumination relationship data, and constructing a second illumination change model based on this; verifying whether the optical physical laws of the first illumination change model and the second illumination change model are compatible in the same time sequence; if the correlation coefficient of the change trends of the two is greater than a preset illumination correlation threshold, then the cross-validation based on the illumination relationship data is determined to be successful.

[0056] Specifically, illumination relationship data is used to characterize the properties of ambient light during evidence collection, such as the direction and intensity of the light source and the pattern of shadow changes. The implementation process is as follows: The server analyzes the videos from the first and second cameras to estimate the illumination parameters of their respective environments. This may include: analyzing the direction and length of shadows cast by objects in the image to infer the direction of the main light source (e.g., the sun's position); calculating the average brightness of the image or the brightness value of a specific area to assess light intensity; and tracking the movement and changes of highlights or shadows over time. Based on the illumination relationship data obtained from the above analysis, a first illumination change model and a second illumination change model are constructed to quantitatively describe the illumination conditions under their respective perspectives. During verification, the focus is on evaluating whether the two models are compatible with the optical physical laws at the same time interval. For example, whether the solar altitude angle change curves inferred by the two models are synchronized, or whether the shadow length change conforms to the time progression of the day. By calculating the correlation coefficient between the two change curves, if the coefficient is greater than a preset illumination correlation threshold, the verification is successful.

[0057] For example, during evidence collection outdoors for several hours, the sun's position constantly changes. A first camera captures the target, while a second camera, though facing a different angle, still reflects ambient light. The server analyzes the first camera video and constructs a model of the changing direction of illumination throughout the day based on changes in the object's shadow. Simultaneously, it analyzes the second camera video and constructs its illumination variation model. By comparing the two models, if the reflected light source (sun) trajectory and light intensity change trends are found to be highly correlated over time (e.g., correlation coefficient > 0.9), the verification is successful, indicating that both videos are under the same natural light environment and have not been affected by localized illumination manipulation.

[0058] In this embodiment, the continuity and physical regularity of ambient lighting are used for verification, which can effectively identify video tampering that violates the laws of natural lighting, such as illogical shadow directions and sudden changes in lighting. This provides support for the continuity of video evidence in the time dimension and its conformity to physical common sense, thereby enhancing the credibility of the evidence.

[0059] According to an embodiment of this application, optionally, the physical feature data includes motion feature data, and the cross-validation mechanism specifically includes: extracting motion vectors, optical flow sequences, or acceleration features of a specific target or global image within the field of view of the first camera based on the video and inertial measurement data in the source dataset of the evidence video to obtain first motion feature data, and constructing a first motion feature model based thereon; extracting motion vectors, optical flow sequences, or acceleration features of a specific target or carrier itself within the field of view of the second camera based on the video and inertial measurement data in the source dataset of the verification video to obtain second motion feature data, and constructing a second motion feature model based thereon; comparing the consistency of the motion trajectory, rhythm, or form described by the first motion feature model and the second motion feature model; if the consistency is higher than a preset motion correlation threshold, then the cross-validation based on the motion feature data is determined to be successful.

[0060] Specifically, motion feature data is used to characterize the motion information of the target or the camera itself in the evidence collection scene, such as motion vectors, optical flow sequences, or acceleration features. The implementation process is as follows: The server combines the video from the first camera with its inertial measurement data to extract motion features of a specific target (such as a moving vehicle) or the global scene within its field of view, obtaining the first motion feature data. This can be achieved, for example, by calculating the optical flow field between consecutive frames, tracking the motion trajectory of a specific target, or fusing IMU data to obtain more accurate motion parameters. Similarly, the data from the second camera undergoes the same processing to extract motion features, obtaining the second motion feature data. Based on these motion feature data, a first motion feature model and a second motion feature model are constructed, respectively. During verification, the consistency of the motion described by the two models is compared. For example, whether the motion trajectory of the vehicle in the first camera video is logically consistent with the motion of the carrier (vehicle) indirectly reflected in the second camera video; or whether the overall optical flow patterns extracted by the two models are compatible. If the consistency index of the motion trajectory, rhythm, or form is higher than a preset motion correlation threshold, the verification is considered successful.

[0061] For example, a patrol car follows another vehicle to collect evidence. A first camera (forward-facing) records the movement of the vehicle in front, while a second camera (inside the vehicle) records the driver's (body swaying due to vehicle acceleration and deceleration). The server extracts the motion vector sequence of the vehicle in front based on the first camera video and the IMU (In-Vehicle Detection Unit), constructing a first motion feature model. Based on the second camera video and the IMU, the server extracts the acceleration features of the driver's body swaying, inferring the movement rhythm of the patrol car itself, and constructing a second motion feature model. Comparison reveals a high degree of consistency between the moment the vehicle in front decelerates and the moment the driver leans forward, indicating a high degree of consistency in movement rhythm. This confirms successful verification, demonstrating that the motion events recorded by the two video feeds are synchronously related.

[0062] In this embodiment, by analyzing the consistency of motion information in the video, the continuity and authenticity of events in dynamic scenes can be effectively verified. This plays an important role in detecting operations that disrupt the continuity of time, such as video frame rate tampering, insertion or deletion of moving objects, and further ensures the reliability of electronic evidence in the spatiotemporal dimension.

[0063] To further improve the accuracy of the verification results, according to embodiments of this application, optionally, the source dataset of the evidence video and the source dataset of the verification video are used together to perform a preset cross-verification mechanism and a benchmark verification mechanism when it is necessary to verify the authenticity of the video captured by the first camera.

[0064] Before executing the cross-validation mechanism, a baseline validation mechanism is also included. This baseline validation mechanism may include the following steps: reconstructing the motion trajectory of the first camera during the second time period of the evidence collection process based on the inertial measurement data, positioning data, and clock data in the source dataset of the evidence collection video from the first camera; reconstructing the motion trajectory of the first camera during the second time period of the evidence collection process based on the inertial measurement data, positioning data, clock data, and the relative positional relationship between the second camera and the first camera in the source dataset of the validation video from the second camera; if the two reconstructed motion trajectories meet the consistency requirements, then the baseline validation for the second time period is successful.

[0065] The first time period may be the same as or different from the second time period; this application does not make any specific restrictions on this.

[0066] For example, in the benchmark verification mechanism, the server selects 10 seconds of data from when the vehicle is turning for verification. Path one directly reconstructs the vehicle's trajectory based on the GPS and IMU data of the first camera on the roof. Path two first reconstructs the vehicle's trajectory based on the IMU data of the second camera inside the vehicle (facing the driver) (because the second camera is rigidly connected to the vehicle body), and then superimposes the pre-stored "position offset of the roof camera relative to the vehicle body" to indirectly calculate the trajectory of the first camera on the roof. Comparison shows that the trajectories of the first camera calculated by the two paths are highly consistent in position and orientation, thus the benchmark verification is successful, proving the continuity and authenticity of the sensor data during the evidence collection period.

[0067] In this embodiment, the motion trajectory of the first camera within the same time period is reconstructed based on the source dataset of the first camera itself and the source dataset of the second camera. If the two reconstruction results meet the consistency requirements, the authenticity of the data collected by the first camera can be verified from the perspective of physical motion, thus eliminating the risk of underlying data failure caused by serious deviation or distortion of data from a single sensor.

[0068] Benchmark verification prioritizes the reliability of motion trajectory data, while cross-verification further validates the consistency of the environmental physical feature model. Through the synergy of benchmark and cross-verification, the accuracy of verification is effectively improved under this dual-verification mechanism, and the system's adaptability to complex scenarios such as violent carrier movement and temporary sensor failures is enhanced. Thus, even if the extraction of physical features required for cross-verification encounters difficulties in certain environments, if benchmark verification passes, it can still prove the basic authenticity of the video source within that time period to a certain extent. Conversely, if benchmark verification fails, it can directly warn of fundamental problems in the data for that time period. This mechanism improves the robustness of the forensic system in non-ideal forensic environments and the credibility of the conclusions.

[0069] According to an embodiment of this application, optionally, a benchmark verification mechanism can be first performed based on the source dataset of the evidence video and the source dataset of the verification video. If the benchmark verification is successful, then a cross-verification mechanism is performed. If the benchmark verification fails, it indicates that the authenticity of the video captured by the first camera is insufficient. If both the benchmark verification and the cross-verification are successful, it indicates that the video captured by the first camera has authenticity.

[0070] Thus, this phased, bottom-up verification architecture avoids potential misjudgments that might occur when high-level feature comparisons are performed directly when the underlying data is unreliable. This systematically reduces the uncertainty of the overall verification process and improves the accuracy of the verification conclusions.

[0071] According to an embodiment of this application, optionally, before performing benchmark verification, the source datasets from the first and second cameras can be time-aligned. High-precision clock data is used to unify inertial measurement unit (IMU) data and positioning data (such as GPS) to the same timestamp sequence, ensuring a consistent time base for subsequent trajectory calculations. The system compares the motion trajectories of the first camera reconstructed from the two paths. "Meeting consistency requirements" typically means that the differences between the two trajectories in key parameters such as position, velocity, and attitude are within a preset error range (e.g., position error less than 1 meter, attitude error less than 3 degrees). If consistent, the benchmark verification is successful, indicating that the sensor data highly matches the physical constraints and the data is highly reliable. If inconsistent, it indicates potential problems such as sensor data forgery, clock asynchrony, or incorrect relative positional relationships.

[0072] According to an embodiment of this application, optionally, before executing the cross-validation mechanism, a benchmark verification mechanism is further included. The benchmark verification mechanism includes: using the extended Kalman filter (EKF) algorithm to calculate the trajectory of the first camera in the world coordinate system based on the inertial measurement data, positioning data, and clock data in the source dataset of the evidence video from the first camera, to determine the first motion trajectory of the first camera in the world coordinate system during the second time period; using the EKF algorithm to calculate the trajectory of the second camera in the world coordinate system based on the inertial measurement data, positioning data, and clock data in the source dataset of the verification video from the second camera, to determine the trajectory of the second camera in the world coordinate system during the second time period; based on the world coordinate system trajectory of the second camera and the relative positional relationship between the second camera and the first camera that is pre-determined or acquired in real time, to determine the second motion trajectory of the first camera in the world coordinate system during the second time period; determining the three-dimensional spatial position deviation between the first motion trajectory and the second motion trajectory at a series of corresponding timestamps during the second time period; if the root mean square value of the spatial position deviation is less than a preset distance threshold, then the benchmark verification during the second time period is determined to be successful.

[0073] Path 1 (Direct Reconstruction): The server constructs an EKF using the IMU data of the first camera itself as control input and GPS data as observations. The EKF predicts the next state of the camera through a motion model, such as a constant velocity model or a constant steering model, and then uses the absolute position observations of GPS to correct this prediction, finally outputting a smooth and continuous first motion trajectory of the first camera in the world coordinate system.

[0074] Path Two (Indirect Reconstruction): This path consists of two sub-steps. First, using the same EKF method as Path One, but employing IMU and GPS data from the second camera, the world coordinate system trajectory of the second camera is calculated. Then, utilizing the pre-determined or real-time acquired relative positional relationship between the second and first cameras—which can be a fixed three-dimensional spatial transformation matrix (including rotation and translation)—this relationship matrix is ​​applied to the trajectory of the second camera, indirectly deriving the second motion trajectory of the first camera.

[0075] The server then aligns the two trajectories on the same timestamp sequence and calculates the Euclidean distance deviation between the two 3D coordinates at each corresponding time point, as well as the root mean square error (RMSE) of the Euclidean distance deviation. RMSE comprehensively reflects the overall deviation of the two trajectories throughout the second time period. If the RMSE value is less than a preset distance threshold (e.g., 0.5 meters), it indicates that the two trajectories are at the same height, and the baseline verification is successful; otherwise, the verification fails, indicating that the data may be abnormal.

[0076] For example, a patrol car serves as a mobile platform for law enforcement recording along a stretch of road. A first camera (primary evidence-gathering camera) is mounted on the roof facing the road, and a second camera (verification camera) is mounted near the rearview mirror inside the vehicle, facing the driver. The relative positions of the two cameras are precisely determined during installation.

[0077] Evidence collection process: The vehicle was traveling in a straight line and then braked. The first camera recorded the road conditions ahead, and the second camera recorded the driver and the limited external environment seen through the windshield.

[0078] Trajectory reconstruction: Path 1, based on GPS / IMU data from the roof-mounted camera, the EKF reconstructs a smooth straight line, with a decrease in speed at the braking point. Path 2, based on GPS / IMU data from the in-vehicle camera, first reconstructs the vehicle's trajectory, then superimposes a fixed relative position relationship to indirectly deduce the trajectory of the roof-mounted camera.

[0079] Consistency check: Calculate the RMSE of the two trajectories. Since the vehicle is a rigid unit, the motion of the two cameras should be consistent. If the RMSE value is very small (e.g., 0.2 meters), far below the threshold (e.g., 1 meter), the verification is successful, proving that the data from the two sensors were consistent and untampered during the evidence collection period. If someone tampered with the GPS signal of the first camera but its IMU data remained unchanged, it would cause a significant deviation between the two trajectories, and the RMSE would far exceed the threshold, causing the verification to fail.

[0080] In this embodiment, the motion trajectory-based benchmark verification method does not rely on external trusted institutions, but instead utilizes the physical constraints between different sensor data within the system for self-verification. This "self-proof" mechanism enhances the inherent credibility and probative value of electronic evidence in judicial proceedings. It is difficult for tamperers to simultaneously and homogeneously forge all sensor data (IMU, GPS) while maintaining their complex dynamics and spatial relationships. Any inconsistencies will be amplified and identified during trajectory comparison. Therefore, this verification scheme has a strong detection capability against tampering with a single data source, such as forging only GPS coordinates or only video. Using RMSE as the consistency criterion ensures that the verification result is no longer a subjective judgment, but a repeatable and quantifiable objective indicator, meeting the standard requirements of objectivity and reliability for judicial evidence.

[0081] Optionally, the evidence collection scenarios in this application may include static scenarios and dynamic scenarios.

[0082] In a static scenario, the relative positions of at least two cameras remain constant. The movable carrier is, for example, a vehicle body, and the first and second cameras are fixed vehicle-mounted cameras. The relative spatial positions of at least two cameras are fixed. In this case, the relative positional relationship between them (such as translation and rotation vectors in three-dimensional space) can be stored as a fixed parameter in a secure storage area on a server or device. This allows for the reconstruction of the first camera's motion trajectory during the first time period of the evidence collection process based on this relative positional relationship during the benchmark verification phase.

[0083] For example, if the mobile platform is a patrol car, during the patrol, a first camera fixedly mounted on the roof continuously records the traffic conditions ahead (evidence video). Simultaneously, a second camera installed in the driver's cab, facing the driver's face, records the driver's liveness (verification video). IMU and GPS data from both cameras are recorded. Since the cameras are rigidly fixed to the vehicle body, their relative positions are fixed and can be predetermined and stored. All data is initially encrypted and signed within the onboard TEE before being transmitted to the command center server via a 5G network.

[0084] In dynamic scenarios, the relative positions between at least two cameras are dynamically changing. For example, if the movable carrier is a wearable device, at least two cameras are installed in different positions on the wearable device, such as different body parts (e.g., helmet, chest armor). Since each body part is movable, the relative positional relationship between the at least two cameras changes continuously as the body parts move. In this case, the relative positional relationship between the two cameras can be obtained in real time, so that the benchmark verification mechanism can reconstruct the motion trajectory of the first camera in the first time period during the evidence collection process based on the relative positional relationship.

[0085] For example, in dynamic scenarios, when a mobile platform is in motion, such as during vehicle-mounted patrols or when law enforcement officers are using wearable devices for mobile evidence collection, if cameras are installed on different moving parts, such as helmets and shoulder armor, their relative positions may change dynamically. In this case, it is necessary to obtain the relative positional relationship in real time. This can be achieved by adding additional positioning modules or visual markers to the platform to obtain the pose changes between the cameras in real time.

[0086] Optionally, according to an embodiment of this application, the method may further include: sending the source dataset of the evidence video and the source dataset of the verification video to a timestamp authentication system, so that the timestamp authentication system initiates a timestamp authentication process on the received dataset and generates a video file and a timestamp certificate for verifying the authenticity and validity of the evidence video.

[0087] Specifically, the forensic system can send the source dataset or its generated digital digest to a standardized time service center (TSA) or a qualified third-party time stamp service. This agency will bind a legally valid and precise authoritative time point certificate to the data packet, generating a digital certificate containing the original data (or its hash) and a timestamp. This process ensures that the data existed before a specific point in time and has not been tampered with.

[0088] To further enhance credibility, according to embodiments of this application, optionally, the method may also include storing the video file and timestamp certificate used to verify the authenticity and validity of the evidence-gathering video in a blockchain system.

[0089] Specifically, leveraging the immutability and distributed ledger characteristics of blockchain, once information is recorded, it cannot be arbitrarily modified by any single party, thus providing reliable integrity assurance for evidence. In judicial proceedings, the integrity of electronic evidence from the date of its storage can be verified by comparing the hash value of the original document with the hash value stored on the blockchain.

[0090] According to an embodiment of this application, optionally, the multi-camera mobile evidence collection system further includes at least one terminal device, which is used to receive video data, inertial measurement data and positioning data from at least two cameras and send them to the server in real time; the at least one terminal device is also used to receive a timestamp certificate.

[0091] Specifically, at least one terminal device, such as an industrial-grade tablet, vehicle-mounted host, or dedicated embedded processing unit, is equipped on a mobile evidence-gathering vehicle like a patrol car or drone. It connects to multiple cameras on the vehicle via wired or wireless means to receive real-time video feeds, inertial measurement unit (IMU) data, and GPS positioning data. The terminal device can locally and temporarily cache and encapsulate the data, and reliably transmit it to a remote server via a wireless network. Furthermore, the terminal device also receives processing results from the server, such as timestamp certificates, and may have a local display interface for on-site personnel to view.

[0092] For example, on a mobile notary vehicle, the terminal device might be an industrial computer installed inside the vehicle, connected to a first camera on the roof for capturing the external environment, a second camera inside the vehicle for recording the notary's procedures, and the vehicle's GPS and IMU modules. The terminal device synchronizes and packages all these data streams, then transmits them to the notary office's cloud server via a 5GCPE device on the roof. Simultaneously, once the server completes timestamp authentication, the terminal device receives and stores the returned timestamp certificate, allowing on-site personnel to confirm that the notarization process has been fully completed.

[0093] In this embodiment, by introducing authoritative timestamps and blockchain evidence storage, internal technical verification (physical consistency) is combined with external legal trust infrastructure, thus constructing a dual guarantee of technical credibility and legal validity, and improving the admissibility and probative value of electronic evidence in judicial proceedings.

[0094] According to embodiments of this application, optionally, at least one of the following is also included: During the evidence collection process, the posture of the first camera remains unchanged for one or more time periods. During the evidence collection process, the second camera maintains its own posture for one or more time periods.

[0095] Specifically, maintaining a constant attitude means that the camera's optical axis (pitch, yaw, rotation angle) remains fixed during evidence collection. This can be achieved by rigidly fixing the camera to a carrier (such as a car roof or tripod) and locking all rotational degrees of freedom of its gimbal. In this mode, since the camera itself does not actively move, the reconstruction of its motion trajectory (in the benchmark verification mechanism) will mainly reflect the motion of the carrier itself. This simplifies the motion analysis model and is beneficial for long-term stable monitoring of stationary or slowly moving targets, or when it is necessary to accurately compare subtle changes in the same scene captured at different time points.

[0096] For example, when conducting targeted monitoring and evidence collection at a site suspected of illegal dumping, law enforcement officers hover a drone equipped with dual cameras at a specific location in the air and lock the gimbal. The first camera (main camera) continuously captures an overhead view of the entire landfill area; the second camera (auxiliary camera) is pointed towards the robotic arm or flight indicator light below the drone. Throughout the hovering evidence collection process, the attitude of both cameras remains absolutely unchanged. In subsequent benchmarking, the motion trajectory reconstructed based on the IMU data of the first camera should approximate a point (with only slight drift), and the trajectory of the first camera reconstructed based on the data from the second camera and its relative position should also be highly consistent with it, thus demonstrating the stability of the perspective and the continuity of the acquired video data during the evidence collection process.

[0097] In this embodiment, the camera attitude holding mode is suitable for scenarios involving long-term stable monitoring of stationary or slowly moving targets. It can effectively reduce errors introduced by the camera's own movement, making the verification focus more on the consistency of the carrier's own movement. This improves the reliability of data and the accuracy of verification in evidence collection tasks that require high-precision positioning or continuous observation of fixed targets.

[0098] Corresponding to the method embodiments of this application, this application also provides a multi-camera mobile evidence collection system, which may include at least two cameras and the aforementioned movable carrier. This multi-camera mobile evidence collection system can be used to implement the various processes of the above method embodiments and can achieve the same technical effects. To avoid repetition, it will not be described again here.

[0099] According to embodiments of this application, optionally, Figure 3 This is a schematic diagram of the structure of a multi-camera mobile evidence collection system provided in an embodiment of this application. The system 300 may include at least two cameras 310, a server 320 and a mobile carrier 330. The server may be a server for a trusted authentication application.

[0100] The server 320 may include a data receiving module, a cross-validation module, and a benchmark verification module; the data receiving module is used to receive the source dataset of the evidence-collecting video and the source dataset of the verification video; the cross-validation module is used to execute the cross-validation mechanism; and the benchmark verification module is used to execute the benchmark verification mechanism.

[0101] Corresponding to the method embodiments of this application, this application also provides a vehicle, which has at least two cameras installed inside and outside the vehicle. The vehicle is used to implement the various processes of the above method embodiments and can achieve the same technical effects. To avoid repetition, it will not be described again here.

[0102] Corresponding to the method embodiments of this application, this application also provides an aircraft with at least two cameras installed inside and outside the aircraft. The aircraft is used to implement the various processes of the above method embodiments and can achieve the same technical effects. To avoid repetition, it will not be described again here.

[0103] Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application.

[0104] like Figure 4 As shown, the electronic device 400 includes a memory 401, a processor 402, and a computer program stored in the memory 401 and executable on the processor 402.

[0105] In one example, the processor 402 described above may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0106] Memory 401 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the mobile forensics method based on multi-camera association verification according to the embodiments of the first aspect of this application.

[0107] The processor 402 runs a computer program corresponding to the executable program code by reading the executable program code stored in the memory 401, in order to implement the mobile forensics method based on multi-camera association verification in the first aspect embodiment above.

[0108] In some examples, electronic device 400 may also include communication interface 403 and bus 410. For example, Figure 4 As shown, the memory 401, processor 402, and communication interface 403 are connected through bus 410 and complete communication with each other.

[0109] The communication interface 403 is mainly used to enable communication between various modules, systems, units, and devices in the embodiments of this application. Input devices and output devices can also be connected through the communication interface 403.

[0110] Bus 410 includes hardware, software, or both, that couples components of electronic device 400 together. For example, and not limitingly, bus 410 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-E) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 410 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0111] The electronic device provided in this application embodiment is capable of achieving Figure 1 The various processes implemented by the electronic device in the method embodiment can achieve the same technical effect, and will not be described again here to avoid repetition.

[0112] In conjunction with the mobile forensics method based on multi-camera correlation verification in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement the steps in the above method embodiments.

[0113] In conjunction with the mobile forensics method based on multi-camera correlation verification in the above embodiments, this application embodiment can provide a computer program product to implement it. This (computer) program product is stored in a non-volatile storage medium, and when executed by at least one processor, it implements the steps in the above method embodiments.

[0114] This application also provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled. The processor is used to run programs or instructions to implement the various processes of the above method embodiments and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0115] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0116] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0117] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0118] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or systems. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0119] The aspects of this disclosure have been described above with reference to flowchart illustrations and block diagrams of methods, systems, and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and block diagrams, and combinations of blocks in the flowchart illustrations and block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing system to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing system, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and flowchart illustrations, and combinations of blocks in the block diagrams and flowchart illustrations, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0120] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A mobile forensics method based on multi-camera correlation verification, characterized in that, The method utilizes vehicle-mounted or airborne evidence-gathering equipment, which includes at least a first camera and a second camera; the method includes: When collecting evidence using vehicle-mounted or airborne evidence-gathering equipment, the first camera continuously captures images of the target object or target location, while the second camera continuously captures images of the internal or external environment of the carrier. The video captured by the first camera, along with the local inertial measurement data and positioning data of the first camera, is used as the source dataset for the evidence video. The video captured by the second camera, along with the local inertial measurement data and positioning data of the second camera, is used as the source dataset for the verification video. The source datasets of the evidence video and the verification video are used together to execute a preset cross-verification mechanism when it is necessary to verify the authenticity and validity of the video captured by the first camera. The cross-verification mechanism includes: Identify at least one set of corresponding physical feature data that exists in the source dataset of the evidence-collecting video and the source dataset of the verification video; Based on the at least one set of corresponding physical feature data, reconstruct the physical feature model of the environment where the first camera is located during the first time period in the evidence collection process; Based on the at least one set of corresponding physical feature data, reconstruct the physical feature model of the environment where the second camera is located during the first time period in the evidence collection process; If the two reconstructed physical feature models meet the requirement of mutual verification, then the cross-validation of the physical feature data for the first time period is successful.

2. The method according to claim 1, characterized in that, Before performing the cross-validation mechanism, the method further includes performing a benchmark validation mechanism, the benchmark validation mechanism comprising: Based on the inertial measurement data, positioning data, and clock data in the source dataset of the evidence-collecting video from the first camera, the motion trajectory of the first camera during the second time period in the evidence-collecting process is reconstructed. Based on the inertial measurement data, positioning data, clock data, and the relative positional relationship between the second camera and the first camera in the source dataset of the verification video from the second camera, the motion trajectory of the first camera during the second time period in the evidence collection process is reconstructed. If the two restored motion trajectories meet the consistency requirements, then the benchmark verification for the second time period is successful.

3. The method according to claim 1, characterized in that, The vehicle-mounted or airborne evidence collection equipment includes evidence collection equipment mounted on any of the following: manned vehicles, unmanned vehicles, manned aircraft, drones, robots, or embodied intelligent agents.

4. The method according to claim 1, characterized in that, The camera includes at least one of the following: monitoring equipment for the interior space of a vehicle, monitoring equipment for people inside a vehicle, monitoring equipment for the exterior space of a vehicle, monitoring equipment for people outside a vehicle, a dashcam, an airborne monitoring device, or a monitoring device mounted on an embodied intelligent agent.

5. The method according to any one of claims 1-4, characterized in that, The physical feature data includes at least one of the following: spatial relationship data, illumination relationship data, audio signal data, motion feature data, and scene feature data.

6. The method according to claim 5, characterized in that, The physical feature data includes spatial relationship data, and the cross-validation mechanism includes: Based on the video in the source dataset of the evidence-collecting video, the three-dimensional coordinates of feature points and the relative spatial positions between objects in the environment captured by the first camera are extracted to obtain the first spatial relationship data, and a first spatial relationship model is constructed based on the first spatial relationship data. Based on the video in the source dataset of the verification video, the three-dimensional coordinates of feature points in the environment where the second camera is located and the relative spatial positions between objects are extracted to obtain the second spatial relationship data, and a second spatial relationship model is constructed based on the second spatial relationship data. The first spatial relationship model and the second spatial relationship model are transformed to the same coordinate system to verify whether their spatial geometric features meet the preset consistency constraints. If the consistency of the spatial geometric features is higher than a preset spatial correlation threshold, then the cross-validation based on spatial relationship data is considered successful.

7. The method according to claim 5, characterized in that, The physical feature data includes audio signal data, and the cross-validation mechanism includes: Based on the audio signal data contained in the source dataset of the evidence-collecting video, a first audio feature model is constructed to reflect the acoustic environment in which the first camera is located. Based on the audio signal data contained in the source dataset of the verification video, a second audio feature model is constructed to reflect the acoustic environment in which the second camera is located. By determining the cross-correlation between the first audio feature model and the second audio feature model, it is verified whether the first audio feature model and the second audio feature model corroborate each other. If the cross-correlation is greater than a preset audio correlation threshold, then the cross-validation based on the audio signal data is considered successful.

8. The method according to claim 5, characterized in that, The physical feature data includes scene feature data, and the static visual features include key points, edge information, or texture feature points. The cross-validation mechanism includes: Based on the videos in the source dataset of the evidence-collecting videos, the static visual features of the shooting scene corresponding to the first camera are extracted to obtain the first scene feature data, and a first scene feature model is constructed based on the first scene feature data. Based on the videos in the source dataset of the verification videos, the static visual features of the shooting scene corresponding to the second camera are extracted to obtain the second scene feature data, and a second scene feature model is constructed based on the second scene feature data. Determine the matching degree of corresponding feature points in the first scene feature model and the second scene feature model; If the matching degree is higher than the preset matching degree threshold, then the cross-validation based on the scene feature data is considered successful.

9. The method according to claim 1 or 2, characterized in that, Before performing the cross-validation mechanism, the method further includes performing a benchmark validation mechanism, the benchmark validation mechanism comprising: Based on the inertial measurement data, positioning data, and clock data in the source dataset of the evidence video from the first camera, the extended Kalman filter (EKF) algorithm is used to calculate the trajectory and determine the first motion trajectory of the first camera in the world coordinate system during the second time period. Based on the inertial measurement data, positioning data, and clock data in the source dataset of the verification video from the second camera, the EKF algorithm is used to calculate the trajectory and determine the world coordinate system trajectory of the second camera in the second time period. Based on the world coordinate system trajectory of the second camera and the relative positional relationship between the second camera and the first camera that is pre-determined or acquired in real time, the second motion trajectory of the first camera in the world coordinate system is determined. Determine the three-dimensional spatial positional deviations of the first motion trajectory and the second motion trajectory at a series of corresponding timestamps within the second time period; If the root mean square value of the spatial position deviation is less than the preset distance threshold, the benchmark verification in the second time period is determined to be successful.

10. The method according to claim 1, characterized in that, Also includes: The source dataset of the evidence video and the source dataset of the verification video are sent to the timestamp authentication system, so that the timestamp authentication system initiates a timestamp authentication process on the received dataset and generates a video file and a timestamp certificate for verifying the authenticity and validity of the evidence video.

11. The method according to claim 10, characterized in that, Also includes: The video file and timestamp certificate used to verify the authenticity of the evidence video are stored in the blockchain system.

12. A multi-camera mobile evidence collection system, characterized in that, The system includes at least two cameras and a movable carrier. The multi-camera mobile evidence collection system is used to implement the mobile evidence collection method based on multi-camera association verification as described in claims 1-11.

13. A vehicle, characterized in that, The vehicle is equipped with at least two cameras inside and outside, and the vehicle is used to implement the mobile forensics method based on multi-camera association verification as described in any one of claims 1-11.

14. An aircraft, characterized in that, The aircraft is equipped with at least two cameras inside and outside, and the aircraft is used to implement the mobile forensics method based on multi-camera correlation verification as described in any one of claims 1-11.

Images