Vehicle-mounted video multi-channel real-time playing and historical playback management system

Abstract

The invention discloses a vehicle-mounted video multi-channel real-time playing and historical playback management system, and relates to the technical field of vehicle-mounted video monitoring, and the vehicle-mounted video multi-channel real-time playing and historical playback management system comprises a double-source modeling module, an event windowing module, a time self-adaption module and a playback positioning module. The double-source modeling module obtains videos in front and in a vehicle, operates an auxiliary algorithm, and constructs a time axis and an index; the event windowing module constructs a time window by taking event triggering as a center, and generates a cross-channel index; the time adaptive module dynamically adjusts a time window according to the vehicle speed, the confidence coefficient and the like, and time axis fragmentation is avoided; and the playback positioning module retrieves video segments based on events to realize multi-channel synchronous playback and rapid positioning, and the system solves the problems of multi-channel video synchronous playing and event accurate positioning, and improves the accident replay efficiency and the safety management level.

Description

Technical Field

[0001] This invention relates to the field of vehicle-mounted video surveillance technology, and more specifically, to a vehicle-mounted video multi-channel real-time playback and historical playback management system. Background Technology

[0002] With the widespread adoption of intelligent vehicle monitoring systems, vehicles are typically equipped with multiple cameras, such as forward-facing ADAS cameras and in-vehicle DSM cameras, for driver assistance and safety monitoring. These systems generate massive amounts of video data in real time and use algorithms to identify various risk events, such as forward collision warnings and abnormal driver behavior. In actual safety management and accident review, effectively and synchronously replaying multi-channel video and quickly locating the video segment corresponding to a specific event has become a key technical challenge.

[0003] In existing technologies, multi-channel video playback typically employs simple, independent timeline management. The lack of precise time synchronization mechanisms between video streams leads to asynchronicity during playback, making it difficult to fully reconstruct the entire event. Furthermore, the correlation between event alarms and video segments is weak, often relying on manual searching of corresponding videos based on alarm timestamps, which is inefficient and prone to errors. Regarding event video segment extraction, most systems use fixed-length extraction methods, failing to dynamically adjust the evidence collection window based on event type, vehicle speed, confidence level, etc., potentially leading to the omission of critical information or storage redundancy. Simultaneously, video segments generated by consecutive alarm events tend to overlap or fragment on the timeline, further increasing the complexity of retrieval and management.

[0004] To address the above problems, this invention proposes a solution. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an in-vehicle video multi-channel real-time playback and historical playback management system to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The vehicle-mounted video multi-channel real-time playback and historical playback management system includes: a dual-source modeling module, an event windowing module, a time-adaptive module, and a playback positioning module, with signal connections between each module; Dual-source modeling module: Acquires frontal image video and in-vehicle image video respectively, runs front assist algorithm and driver behavior algorithm respectively, establishes event category and parameter configuration table, forms parameters such as detection duration, start speed, alarm interval, baseline expansion coefficient, speed sensitivity coefficient and algorithm confidence, and constructs ordinary time axis and initial frame level index; Event windowing module: Constructs a basic time window centered on the event trigger time, introduces an expansion coefficient to calculate the start and end times of a unified time interval, obtains the event time axis by discretizing the ordinary time axis, and generates cross-channel index records according to event type and channel weight, and associates the unified time interval with the frame-level index of each video channel; Time Adaptive Module: On the event timeline, the module dynamically adjusts the expansion coefficient and time window length using the vehicle's current speed, algorithm confidence, and alarm interval parameters. This allows the time window to adapt to the event category and vehicle speed. It also merges or trims the time intervals of adjacent events in the same channel to avoid fragmentation of the timeline and outputs a unified time interval and event timeline after adaptive adjustment. Playback Positioning Module: Based on the upper-layer event positioning request, retrieve the target event record, determine the overall time interval and associated channel set, map the time interval to the frame-level position of each channel, generate multi-channel playback tasks and send them to the terminal. The platform controls the synchronous playback of multiple channels according to the time scale and time tolerance, and realizes fast jump and accurate positioning of single or multiple events on the time axis through playback progress control.

[0007] In a preferred embodiment, the dual-source modeling module includes the following steps: The system acquires images and videos of the front of the vehicle captured by the forward-looking ADAS camera and images and videos of the interior of the vehicle captured by the in-vehicle DSM camera. The system uses a forward assist algorithm for lane line recognition, vehicle tracking and distance estimation, and human target recognition, and a driver behavior algorithm for face detection and behavior recognition. When the warning conditions are identified, the detection results are structured and encapsulated according to the event category and parameter configuration table to generate a single detection output containing event type identifier, event trigger time, detection duration, algorithm confidence, current vehicle speed and video channel information; Based on the system clock and video encoding frame rate, the vehicle running time is continuously divided into equally spaced time scales, and a global timestamp is assigned to each time scale. Record the time scale index and frame position in the storage medium corresponding to all uploaded video frames to form an initial frame-level index, with the ordinary time axis serving as the time scale reference; The structure descriptions of the event categories and parameter configuration tables, single detection output formats, ordinary timelines, and frame-level indexes are synchronized to the platform side. The platform side records the parameter version number to distinguish different versions of event records, ensuring the consistency and traceability of the event processing process.

[0008] In a preferred embodiment, the event windowing module includes the following steps: Based on the event trigger time Centered on the duration of the detection Construct a basic time window with start and end times of [time value]. , ; Based on the baseline expansion factor Algorithm alarm interval The expansion factor is calculated using the following formula: Calculate the actual expansion factor to obtain the start and end times of the unified time interval: , ; The event timeline is discretized based on the sampling interval of the ordinary timeline, and a cross-channel index record containing the event number, event type, unified time interval, associated channel set and channel weight is generated for each event. The channel weight is calculated and determined based on the video channel associated with the event type and the risk contribution. The cross-channel index record is used to establish a mapping relationship between the event timeline and the frame-level index of each video channel.

[0009] In a preferred embodiment, the time-adaptive module includes the following steps: Based on event category c, current vehicle speed v, event confidence level ρ, and starting speed, the expansion coefficient is calculated using the following formula: ;in, Event Category The benchmark coefficient, It is the starting speed set by the algorithm in the dual-source modeling module. Used to adjust the effect of vehicle speed on the length of the time window. It is used to correct the expansion coefficient based on the confidence level, dynamically adjust the expansion coefficient k, and thus adjust the length of the unified time interval; For consecutive alarm events occurring on the same channel, if the trigger time interval is less than the alarm interval set by the algorithm... Then the unified time intervals corresponding to the alarm events will be merged; When the detection time returned by the algorithm is an interval, the time-adaptive module adjusts the time based on the real-time detected action duration. ,pass Determine the detection duration used to construct the time window.

[0010] In a preferred embodiment, the playback positioning module includes the following steps: Parse the event location request to obtain the target event identifier set and its associated device; Based on the target event identifier, the corresponding record is retrieved from the cross-channel index table to obtain a unified time interval and a set of associated channels. When multiple events are involved, the union of their unified time intervals is calculated to obtain the overall time interval. ; The overall time interval is discretized into a time scale sequence according to the sampling period of the ordinary time axis, and each time scale is mapped to the video frame position of each channel using frame-level index information; Send a playback control command containing the total time interval and channel set to the terminal; The video streams from each channel transmitted back by the receiving terminal are calibrated and buffered according to the time scale-frame position index and time tolerance parameters to achieve multi-channel synchronous playback; Based on the event positions marked on the event timeline, the playback progress control module enables quick jumps and precise positioning for single or multiple events.

[0011] The technical effects and advantages of the vehicle-mounted video multi-channel real-time playback and historical playback management system of the present invention are as follows: This invention fundamentally solves the problem of asynchronous playback during multi-channel playback by constructing a unified general timeline and event timeline and precisely associating the frame-level indexes of multiple video streams with them, ensuring the authenticity and integrity of event playback. The event windowing module generates event records containing unified time intervals and cross-channel indexes, enabling upper-layer applications to directly retrieve data based on event identifiers. The playback positioning module can quickly map logical time intervals to specific frame positions in each channel, achieving second-level rapid jumps and precise positioning of single or multiple events, greatly simplifying the operation process. The time-adaptive module can dynamically adjust the length and expansion coefficient of the evidence collection window based on multiple dimensions such as event category, real-time vehicle speed, and algorithm confidence, making the video capture range more closely match the actual risk level, avoiding the loss of key information and reducing the storage of invalid data. By intelligently merging or cropping the time intervals of adjacent events in the same channel according to the alarm interval, fragmented intervals on the timeline are effectively reduced, maintaining the continuity and neatness of the timeline, facilitating macro-level browsing and continuous playback. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the in-vehicle video multi-channel real-time playback and historical playback management system module of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Example Please see Figure 1As shown, the present invention discloses an in-vehicle video multi-channel real-time playback and historical playback management system, including: a dual-source modeling module, an event windowing module, a time-adaptive module, and a playback positioning module, with signal connections between the modules; Dual-source modeling module: Acquires frontal image video and in-vehicle image video respectively, runs front assist algorithm and driver behavior algorithm respectively, establishes event category and parameter configuration table, forms parameters such as detection duration, start speed, alarm interval, baseline expansion coefficient, speed sensitivity coefficient and algorithm confidence, and constructs ordinary time axis and initial frame level index; Event windowing module: Constructs a basic time window centered on the event trigger time, introduces an expansion coefficient to calculate the start and end times of a unified time interval, obtains the event time axis by discretizing the ordinary time axis, and generates cross-channel index records according to event type and channel weight, and associates the unified time interval with the frame-level index of each video channel; Time Adaptive Module: On the event timeline, the module dynamically adjusts the expansion coefficient and time window length using the vehicle's current speed, algorithm confidence, and alarm interval parameters. This allows the time window to adapt to the event category and vehicle speed. It also merges or trims the time intervals of adjacent events in the same channel to avoid fragmentation of the timeline and outputs a unified time interval and event timeline after adaptive adjustment. Playback Positioning Module: Based on the upper-layer event positioning request, retrieve the target event record, determine the overall time interval and associated channel set, map the time interval to the frame-level position of each channel, generate multi-channel playback tasks and send them to the terminal. The platform controls the synchronous playback of multiple channels according to the time scale and time tolerance, and realizes fast jump and accurate positioning of single or multiple events on the time axis through playback progress control.

[0015] In the dual-source modeling module, forward-facing video and in-vehicle video are acquired separately. Forward assist algorithms and driver behavior algorithms are run separately, and an event category and parameter configuration table is established. This generates parameters such as detection duration, vehicle speed, alarm interval, baseline expansion coefficient, speed sensitivity coefficient, and algorithm confidence. A standard timeline and initial frame-level index are constructed. Specific content includes: The system acquires video images of the vehicle's front from the forward-facing ADAS camera and video images of the driver and cab from the in-vehicle DSM camera, with both videos encoded and stored simultaneously. The ADAS camera continuously performs lane line recognition, vehicle tracking and distance estimation, and human target recognition on the forward-facing images. The DSM camera is equipped with face detection and behavior recognition models to detect states such as closed eyes, yawning, smoking, making phone calls, and distraction. When the Advanced Driver Assistance System (ADAS) or Driver Behavior Monitoring System (DSM) identifies a warning condition, the dual-source modeling module encapsulates the detection results in a structured manner according to the event category configuration table, generating a single detection output, which includes at least: an event type identifier. Event triggering time The detection duration output by the algorithm Algorithm confidence Current vehicle speed The corresponding video channel information and the detection duration are based on the standard parameters of the product algorithm. For example, it can be set to 3 seconds for FCW detection, 1 second for HMW detection, instantaneous for LDW detection, 5-30 seconds for smoking detection in DSM, 3 seconds for distraction detection, 3 seconds for closed eyes detection, 10 seconds for phone call detection, and 20 seconds for abnormal driver detection. In addition to detection time, each algorithm is also configured with alarm interval and start speed to avoid false alarms and repeated alarms. When an event occurs, the terminal caches the above parameters together with the algorithm confidence and synchronizes them to the platform database. To achieve multi-channel synchronization, a normal timeline is constructed. Based on the system clock and video frame rate, the vehicle running time is continuously divided into equally spaced scales. Assuming the terminal video encoding frame rate is 25 fps, one second contains 25 scales. Each scale records a global timestamp. The arrival time of all uploaded video frames is marked, and the timeline is discretized according to the sampling period. This normal timeline does not involve events or cross-channel indexes; it only lays the foundation for synchronized playback. The platform supports simultaneous operation of two algorithms, requiring independent caching for each channel when receiving video data. Each time point recorded on the ordinary timeline is linked to the original video frames of different channels. Through buffer management and timestamp alignment, all frames originate from the same system time. The ordinary timeline and frame-level index are initialized. In order to achieve multi-channel synchronous playback, the dual-source modeling module constructs an ordinary timeline based on the terminal system clock and video encoding frame rate. The vehicle running time is divided into equally spaced time scales, and a global timestamp is assigned to each time scale. For all uploaded video frames, the corresponding time scale index and the frame sequence number or file offset in the local storage medium are recorded to form the initial frame-level index of time scale-channel-frame position. In this step, the ordinary timeline is only used as a time scale reference and does not contain event information. The dual-source modeling module synchronizes the structural descriptions of event categories and parameter configuration tables, single-detection output formats, ordinary timelines, and frame-level indexes to the platform side in the form of configuration data and runtime status. The platform side records the parameter version number. When the parameter table is modified, the version number is used to distinguish between old and new event records. This ensures that the event windowing module, time adaptive module, and playback positioning module can all perform calculations and interpretations according to the parameter version corresponding to the event at the time of event generation when performing event windowing, time adaptive adjustment, and playback positioning. This guarantees the consistency and traceability of the timeline construction and event positioning process.

[0016] In the event windowing module, a basic time window is constructed centered on the event trigger time. An expansion coefficient is introduced to calculate the start and end times of a unified time interval. The event time axis is discretized based on the ordinary time axis, and cross-channel index records are generated according to event type and channel weight. The unified time interval is associated with the frame-level index of each video channel. Specific content includes: When a warning condition is detected, the event trigger time is output synchronously. and the duration of the event detection To extract video segments related to the event from the ordinary timeline, a symmetrical time window is constructed centered on the event trigger time, assuming the detection duration is... The basic time window length is The start and end times are , ;in, The detection parameters correspond to the algorithm, for example, the detection time for FCW is 3 seconds, HMW is 1 second, and LDW is considered as an instantaneous event. It can be set to 2 seconds, DSM smoking detection 5-30 seconds, closed-eye detection 3 seconds, and phone call detection 10 seconds; When the detection time returned by the algorithm is an interval or uncertain, the platform determines the duration based on the real-time detected action duration. Select For transient events, such as lane departure or driver abnormality, a default detection time is defined. As a buffer window; To ensure that the scope of evidence collection covers both the precursors and consequences of events, a time window expansion function is introduced on the basic time window. This function automatically calculates the expansion coefficient based on the event detection time and the alarm interval of the algorithm. This serves as the baseline expansion coefficient for a certain type of event. The alarm interval specified by the algorithm, such as 15 seconds for FCW, 15 seconds for HMW, and 20 seconds for smoking detection, is defined by the expansion factor as: ;in, This is an adjustable weighting constant used to balance the relationship between the alarm interval and the detection time. For events with a long alarm interval but a short detection time, such as a driver anomaly detection time of 20 seconds and an alarm interval of 900 seconds, It will automatically increase in size to ensure more comprehensive evidence collection; however, for events with longer detection times, A moderate increase is sufficient; the start and end times of the unified time interval are determined by... , Calculations show that it can be configured on the platform. and To accommodate the risk preferences of different teams; An event timeline is discretized within a unified time interval based on the sampling interval of the ordinary timeline. The ordinary timeline is based on the terminal encoding frame rate. Based on, it generates per second Each scale; the event timeline starts from... arrive Enumerate each time scale in turn , ,in Each scale mark records absolute time. The event identifier and reserved channel fields are used to map these scales to the corresponding video frame indexes of each video channel using a data structure, forming an event timeline. This timeline supports playback positioning and can also serve as a synchronization benchmark for real-time playback. To quickly locate video frames from different channels in subsequent scheduling, a cross-channel index record is built for each event. The index record includes: event number, event type, and start and end times of a unified time interval. Related channel set Channel weight ; The channel set and weights are derived from the application scope of the algorithm in the dual-source modeling module: forward collision, lane departure, close following distance, and pedestrian detection are mainly associated with the forward-looking channel; smoking, making phone calls, distraction, and closing eyes are mainly associated with the in-vehicle channel; Collisions may involve multiple channels, therefore weighting can be calculated based on risk contribution, such as... The index table maps the event timeline to the frame number of each channel, providing a precise mapping relationship for subsequent real-time scheduling and historical retrieval; The originally scattered alarm events were converted into a unified time interval and linked with the ordinary time axis through the event time axis. The cross-channel index establishes a mapping between events and video channels, laying the data foundation for subsequent real-time scheduling and historical queries.

[0017] In the adaptive time module, the expansion coefficient and time window length are dynamically adjusted on the event timeline using the vehicle's current speed, algorithm confidence, and alarm interval parameters. This allows the time window to adapt to different event categories and vehicle speeds. Time intervals of adjacent events in the same channel are merged or cropped to avoid timeline fragmentation. The module outputs a unified time interval and event timeline after adaptive adjustment. Specific details include: Length of uniform time interval The event window module has an expansion coefficient. With detection time Joint decision, that is However, different types of events have different requirements for the length of evidence collection. Therefore, this embodiment dynamically adjusts the length during runtime. Value, let the event category be The vehicle's current speed is The event confidence level is The expansion coefficient can then be calculated using the following function: ;in, Event Category The benchmark coefficient, It is the starting speed set by the algorithm in the dual-source modeling module. Used to adjust the effect of vehicle speed on the length of the time window. Used to adjust the expansion coefficient based on the confidence level, when When the speed exceeds the starting speed, the expansion factor increases to cover a larger risk range at high speeds; when the confidence level is low... Approaching 0, the expansion coefficient is moderately increased, and the time window length is related to both the event category and the real-time speed and algorithm confidence, thus avoiding resource waste or insufficient information caused by a fixed multiple; The detection duration of the algorithm is the basis for constructing the time window. The dual-source modeling module provides the detection time interval for each algorithm; for example, DSM smoking detection is 5–30 seconds, eye-closed detection is 3 seconds, and phone call detection is 10 seconds. This applies when the algorithm returns an uncertain detection time or only provides a range. At that time, the duration of the action captured from the terminal Specific The value is calculated as follows: If the duration of a real-time action exceeds the upper limit set by the algorithm, the upper limit value is used; otherwise, the lower limit value is used to ensure that the time window covers key behaviors without being excessively long. For instantaneous events, the default detection time is used. As ; Vehicle speed directly affects the severity of an incident and the need for evidence collection. To maintain consistency with the algorithm activation threshold in the dual-source modeling module, speed is incorporated into the calculation of the expansion coefficient. Specifically, let the starting speed be... The vehicle's current speed is The velocity factor is then defined as: ;in, For the speed sensitivity coefficient related to the event category, when Time velocity factor Set to 1, maintaining the baseline time window; when When the threshold is exceeded, the speed factor increases linearly with speed, reflecting the need for a wider time window at high speeds. This speed factor is multiplied in the calculation of the expansion coefficient, thus making the time window length adaptively change with vehicle speed, rather than simply using a fixed multiple of a simple segment. To prevent timeline fragmentation caused by continuous alarms, each algorithm sets alarm intervals in the dual-source modeling module, such as a 15-second interval for FCW and a 20-second interval for DSM smoke detection. When adjusting the time window, an interval constraint rule is introduced: for two consecutive alarms occurring on the same channel... and First, calculate their unified time interval independently. , If the distance between the trigger time of the second alarm and the uniform interval of the first alarm is less than the alarm interval set by the algorithm... If they are considered as the same risk process, the two time windows can be merged into one. If there is overlap between time windows but the interval is greater than the alarm interval, the time window of the main event is retained according to the event priority, such as the safety warning of ADAS taking priority over the distraction alarm of DSM. The time window of the secondary event is only retained in the part outside the main event. Through this interval-based merging and trimming, unnecessary overlap can be effectively reduced and the continuity of the time axis can be maintained. The unified time interval can be dynamically changed based on event type, duration, vehicle speed, and alarm interval to better suit actual needs. This process is completed on the platform side and does not affect the real-time performance of the terminal algorithm.

[0018] In the playback positioning module, the target event record is retrieved based on the upper-layer event positioning request, the overall time interval and associated channel set are determined, the time interval is mapped to the frame-level position of each channel, a multi-channel playback task is generated and sent to the terminal, the platform controls the synchronous playback of multiple channels according to the time scale and time tolerance, and realizes fast jumping and precise positioning of single or multiple events on the timeline through playback progress control. Specific content includes: Based on the standard timeline, event timeline, and cross-channel index table, the target event is precisely located, and all video channels are driven to play back synchronously according to a unified timeline. Specifically, this includes: The event location module receives an event location request from the upper layer application. The request includes at least the target event identifier, the target device identifier, and the location mode parameters. The event identifier corresponds one-to-one with the event records generated in the event windowing module and the time adaptive module. The location mode parameters are used to indicate whether to perform single event playback, event sequence playback, or playback according to time range. The event location module parses the request to obtain the set of target event identifiers and their respective devices. In the cross-channel index table, the corresponding index record is retrieved based on the target event identifier. From the index record, the parameter set consistent with the event windowing module and the time adaptive module is read, including: event category and event trigger time. Detection duration Expansion coefficient , unified time interval start time End time of unified time interval Associated channel sets and priorities; When the positioning mode involves multiple events, the event positioning module merges the multiple events based on a unified time interval: for multiple events belonging to the same positioning task on the same device, the union of their unified time intervals is calculated to obtain the overall time interval of this positioning task. And merge all related channels involved in the event into a unified channel set; Based on the standard timeline and event timeline, time intervals are... Perform mapping and discretization: Calculate the time scale sequence within the time interval based on the sampling period of the standard time axis or the video frame rate. This yields the start and end tick indices on the timeline. Using the frame-level index information pre-stored for each channel in the cross-channel index table, each time scale is... Mapped to video frame numbers or file offsets for each channel, resulting in a multi-dimensional index set of time scale-channel-frame position; To eliminate the impact of terminal encoding jitter and network latency, the event localization module aligns the timestamps of each channel, introducing a time tolerance parameter. Frames whose timestamps fall within the tolerance range are considered usable frames of the same scale and are then compared with the time scale. Binding creates an event timeline for synchronous playback; Based on the overall time interval And a unified channel set, generating playback task description information. The task description includes parameters such as device identifier, channel identifier set, start time, end time, and frame-level index range; After receiving the playback task description, the playback scheduling module sends a playback control command to the corresponding terminal. The command carries the start and end times of a unified time interval and the channel set. The terminal reads the video data of each channel in the corresponding time interval from the local storage according to the command, encapsulates it into a playback bitstream according to the timestamp order and sends it back to the platform. The platform performs secondary calibration and buffer arrangement of each channel bitstream according to the aforementioned time scale-frame position index, so that different channels only output synchronized frames after time tolerance correction at the same time scale. During playback, the event location module marks the trigger position of each event and the start and end positions of a unified time interval on the time scale sequence according to the event timeline, and shares the event marking information with the playback progress control module: The playback progress control module determines the relationship between the playback position and the unified time interval of each event in real time based on the current time scale index. When the playback progress enters or leaves the unified time interval of a certain event, a status notification is issued to drive the interface layer or upper layer business to switch states. When the upper layer needs to jump the playback progress to a certain event moment, it only needs to specify the event identifier. The playback progress control module adjusts the current playback position to the time scale corresponding to the event trigger moment or the starting scale of a unified time interval through the time scale index returned by the event positioning module, so as to achieve quick positioning of the event. When the positioning mode is event sequence playback, the playback progress control module jumps to the same time interval of each event in the order recorded in the event timeline, and continuously completes the positioning and playback of multiple events. Through four steps, this invention solves the problems of asynchronous playback of multi-channel videos and difficulty in locating and replaying alarm events in existing technologies, achieving unified timeline construction and cross-channel management centered on event time. The system not only supports parallel operation of ADAS and DSM dual algorithms, but also allows for flexible adjustment of time window strategies on the platform to adapt to the needs of different vehicles and business scenarios. The time alignment and rapid location functions for real-time playback and historical playback significantly improve the efficiency of accident review and the level of safety management. The above embodiments are merely illustrative examples of this technical solution. Those skilled in the art can make adjustments and improvements to the details and parameters without departing from the principles of this invention, and these equivalent changes should also be considered to fall within the protection scope of this invention.

[0019] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0020] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0021] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0022] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0023] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0024] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A vehicle-mounted video multi-channel real-time playback and historical playback management system, characterized in that, Signal connections between modules; Dual-source modeling module: Acquires frontal image video and in-vehicle image video respectively, runs front assist algorithm and driver behavior algorithm respectively, establishes event category and parameter configuration table, forms detection duration, start speed, alarm interval, baseline expansion coefficient, speed sensitivity coefficient and algorithm confidence parameters, and constructs ordinary time axis and initial frame level index; Event windowing module: Constructs a basic time window centered on the event trigger time, introduces an expansion coefficient to calculate the start and end times of a unified time interval, obtains the event time axis by discretizing the ordinary time axis, and generates cross-channel index records according to event type and channel weight, and associates the unified time interval with the frame-level index of each video channel; Time Adaptive Module: On the event timeline, the module dynamically adjusts the expansion coefficient and time window length using the vehicle's current speed, algorithm confidence, and alarm interval parameters. This allows the time window to adapt to the event category and vehicle speed. It also merges or trims the time intervals of adjacent events in the same channel to avoid fragmentation of the timeline and outputs a unified time interval and event timeline after adaptive adjustment. Playback Positioning Module: Based on the upper-layer event positioning request, retrieve the target event record, determine the overall time interval and associated channel set, map the time interval to the frame-level position of each channel, generate multi-channel playback tasks and send them to the terminal. The platform controls the synchronous playback of multiple channels according to the time scale and time tolerance, and realizes fast jump and accurate positioning of single or multiple events on the time axis through playback progress control.

2. The vehicle-mounted video multi-channel real-time playback and historical playback management system according to claim 1, characterized in that, The dual-source modeling module is specifically used for: The system acquires images and videos of the front of the vehicle captured by the forward-facing ADAS camera and images and videos of the interior of the vehicle captured by the in-vehicle DSM camera. The system uses a forward assist algorithm for lane line recognition, vehicle tracking and distance estimation, and human target recognition, and a driver behavior algorithm for face detection and behavior recognition. When a warning condition is identified, the detection result is structured and encapsulated according to the event category and parameter configuration table to generate a single detection output containing event type identifier, event trigger time, detection duration, algorithm confidence, current vehicle speed, and video channel information.

3. The vehicle-mounted video multi-channel real-time playback and historical playback management system according to claim 2, characterized in that, Based on the system clock and video encoding frame rate, the vehicle running time is continuously divided into interval time scales, and a global timestamp is assigned to each time scale. Record the time scale index and frame position in the storage medium corresponding to all uploaded video frames to form an initial frame-level index, with the ordinary time axis serving as the time scale reference.

4. The vehicle-mounted video multi-channel real-time playback and historical playback management system according to claim 2, characterized in that, The structure descriptions of the event categories and parameter configuration tables, single detection output formats, ordinary timelines, and frame-level indexes are synchronized to the platform side. The platform side records the parameter version number to distinguish different versions of event records, ensuring the consistency and traceability of the event processing process.

5. The vehicle-mounted video multi-channel real-time playback and historical playback management system according to claim 1, characterized in that, The event-driven window module is specifically used for: Based on the event trigger time Centered on the duration of the detection Construct a basic time window with start and end times of [time value]. , ; Based on the baseline expansion factor Algorithm alarm interval The expansion factor is calculated using the following formula: Calculate the actual expansion factor to obtain the start and end times of the unified time interval: , ; The event timeline is obtained by discretizing the sampling interval of the ordinary timeline, and a cross-channel index record containing the event number, event type, unified time interval, associated channel set and channel weight is generated for each event.

6. The in-vehicle video multi-channel real-time playback and historical playback management system according to claim 1, characterized in that, The channel weight is calculated and determined based on the video channel associated with the event type and the risk contribution. The cross-channel index record is used to establish a mapping relationship between the event timeline and the frame-level index of each video channel.

7. The in-vehicle video multi-channel real-time playback and historical playback management system according to claim 1, characterized in that, The time-adaptive module is specifically used for: Based on event category c, current vehicle speed v, event confidence level ρ, and starting speed, the expansion coefficient is calculated using the following formula: ;in, Event Category The benchmark coefficient, It is the starting speed set by the algorithm in the dual-source modeling module. Used to adjust the effect of vehicle speed on the length of the time window. It is used to correct the expansion coefficient based on the confidence level, dynamically adjust the expansion coefficient k, and thus adjust the length of the unified time interval; For consecutive alarm events occurring on the same channel, if the trigger time interval is less than the alarm interval set by the algorithm... Then, the unified time intervals corresponding to the alarm events will be merged.

8. The vehicle-mounted video multi-channel real-time playback and historical playback management system according to claim 7, characterized in that, When the detection time returned by the algorithm is an interval, the time-adaptive module adjusts the time based on the real-time detected action duration. ,pass Determine the detection duration used to construct the time window.

9. The vehicle-mounted video multi-channel real-time playback and historical playback management system according to claim 1, characterized in that, The playback positioning module is specifically used for: Parse the event location request to obtain the target event identifier set and its associated device; Based on the target event identifier, the corresponding record is retrieved from the cross-channel index table to obtain a unified time interval and a set of associated channels. When multiple events are involved, the union of their unified time intervals is calculated to obtain the overall time interval. ; The overall time interval is discretized into a time scale sequence according to the sampling period of the ordinary time axis, and each time scale is mapped to the video frame position of each channel using frame-level index information.

10. The vehicle-mounted video multi-channel real-time playback and historical playback management system according to claim 9, characterized in that, Send a playback control command containing the total time interval and channel set to the terminal; The video streams from each channel transmitted back by the receiving terminal are calibrated and buffered according to the time scale-frame position index and time tolerance parameters to achieve multi-channel synchronous playback; Based on the event positions marked on the event timeline, the playback progress control module enables quick jumps and precise positioning for single or multiple events.

Images