Vehicle-mounted multi-device real-time positioning and video linkage system based on Beidou and GPS

Abstract

The application discloses a vehicle-mounted multi-device real-time positioning and video linkage system based on Beidou and GPS, relates to the technical field of vehicle-mounted positioning and video linkage, and is suitable for the scene where a continuous difference in a heading direction is caused by unilateral continuous shielding. The system performs a continuous difference into a gate limiting effective working time window based on a difference sequence of a Beidou heading direction change and a GPS heading direction change after alignment, generates a jump activity index in the window, and combines a lateral acceleration integral trend index to cross-determine a true or false side slip, outputs a false trigger interception signal or a real side slip event anchor point moment, and updates a unilateral compression index in a statistical period, then locks an asymmetric backtracking distribution of a driver backtracking depth and a forward backtracking depth under a total backtracking budget constraint, backtracks a two-way video segment to generate a linkage evidence package for uploading, suppresses overwrite pollution, and guarantees backtracking coverage depth.

Description

Technical Field

[0001] This invention relates to the field of vehicle positioning and video linkage technology, and more specifically, to a vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS. Background Technology

[0002] Vehicle-mounted positioning and video linkage technology is commonly used on the terminal side to continuously collect forward environmental video streams and driver status video streams, and write them to a local circular buffer. This allows for the retrieval of evidence fragments after an event is triggered and uploaded to the platform for linked playback. Existing solutions typically receive BeiDou and GPS positioning outputs and calculate observations such as heading changes and vehicle speed to generate event triggers and mark the trigger time. This is combined with trigger de-jittering and trigger frequency limiting rules, as well as fixed driver backtracking depth and forward backtracking depth configurations to complete the fragment solidification and uploading.

[0003] The existing technology has the following shortcomings: On the one hand, under continuous unilateral obstruction, the BeiDou heading change and GPS heading change may show a continuous difference, and the GPS heading change is prone to short-term spike jumps. This causes the heading observation to be affected by the mixed influence of obstruction error and short-term jumps, making it difficult for triggering decisions and frequency limiting decisions to distinguish between false triggers and real sideslip. The consequence is that the system is either misled into frequent hardening and uploading, leading to rapid cache overwriting and evidence contamination, or the trigger sensitivity is reduced to suppress overwriting, resulting in missed reports and records when real sideslip occurs. The difficulty lies in the fact that when observables are affected by obstruction and there is a lack of executable criteria, a single threshold or simple frequency limiting cannot simultaneously take into account both types of consequences.

[0004] On the other hand, with a fixed ring cache capacity and limited uplink bandwidth, if a fixed driver backtracking depth and forward backtracking depth configuration is still used for a long time, it will be difficult to adjust the backtracking allocation according to pollution pressure and event importance under the constraint of the total backtracking budget during periods of high trigger density. This can easily lead to invalid segments occupying upload density and crowding out the backtracking coverage depth of critical historical segments. The difficulty lies in the fact that resource pressure fluctuates with the statistical period, and there is a linkage constraint between backtracking allocation and trigger density. Relying on static configuration or manual parameter tuning makes it difficult to achieve stable convergence under different road and occlusion conditions.

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a vehicle-mounted multi-device real-time positioning and video linkage system based on Beidou and GPS to solve the problems mentioned in the background art. Summary of the Invention

[0006] To achieve the above objectives, the present invention provides the following technical solution: A vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS includes: The acquisition module is used to acquire forward environment video, driver status video, Beidou heading changes, GPS heading changes, and vehicle inertial measurement unit output information, and write them to the local cache. The difference trigger module is used to enter the authenticity judgment module when the difference between the Beidou heading change and the GPS heading change meets the continuous judgment duration condition. The authenticity discrimination module is used to combine the jump state of GPS heading change with the output information of the inertial measurement unit to distinguish between false triggers and real sideslips. When it is distinguished as a false trigger, it outputs a false trigger interception signal; when it is distinguished as a real sideslip, it outputs the anchor point time of the real sideslip event. The depth allocation module is used to determine the backtracking depth of the driver status video based on the pseudo-trigger interception signals within the statistical period, under the constraint of the total backtracking budget, and then determine the backtracking depth of the forward environment video based on the remaining backtracking budget.

[0007] In a preferred embodiment, the acquisition module generates a unified timestamp for the forward environment video, driver status video stream and sensor samples based on a unified clock source, and performs writing and time alignment based on the unified timestamp. The vehicle sensor samples include at least BeiDou heading changes, GPS heading changes and inertial measurement unit output information.

[0008] In a preferred embodiment, the local circular cache is a timestamp-indexed circular storage structure that maintains at least the write pointer and the mapping relationship between the timestamp and the cache location.

[0009] In a preferred embodiment, the BeiDou heading change and the GPS heading change are obtained from the difference in heading values ​​at adjacent sampling times, and when the vehicle speed is lower than a preset speed threshold, the BeiDou heading change and the GPS heading change do not participate in the subsequent determination; the continuous determination duration condition includes that the absolute value of the difference between the aligned BeiDou heading change and the GPS heading change is not less than a preset difference threshold and the continuous determination duration is continuously reached.

[0010] In a preferred embodiment, the output information of the inertial measurement unit includes at least lateral acceleration information, and the sideslip trend corresponding to the lateral acceleration information can be characterized by a lateral acceleration integral trend index. The lateral acceleration integral trend index is the window integral quantity obtained by integrating the lateral acceleration after zero bias correction within the sliding time window, and the lateral acceleration zero bias estimate is updated when a preset low dynamic condition is met. The sideslip trend condition includes that the lateral acceleration integral trend index is not less than a preset sideslip trend threshold within a preset trend window.

[0011] In a preferred embodiment, the jump state of GPS heading change can be characterized by a jump activity index; the jump activity index is the number of samples in which the absolute value of GPS heading change exceeds a preset jump threshold within a preset sliding time window, and the jump condition includes that the jump activity index is not less than the preset activity threshold.

[0012] In a preferred embodiment, the cross-determination of the authenticity discrimination module satisfies the following rules: when the jump condition is met and the sideslip trend condition is not met, a false trigger interception signal is generated; when the sideslip trend condition is met, the true sideslip event anchor point time is generated; and when both the jump condition and the sideslip trend condition are met, the true sideslip event anchor point time is generated.

[0013] In a preferred embodiment, the false trigger interception signal within the statistical period is used to characterize the degree of change in the driver's status video retrospective demand, and can be represented by the one-sided pressure index in the calculation. The one-sided pressure index is obtained by converting the number of false trigger interception signals within the preset statistical period into the preset statistical period duration. The one-sided pressure index is updated using a sliding statistical window counting method, and when no false trigger interception signal is generated, the one-sided pressure index is kept at the most recently updated value, and when there is no most recently updated value, the preset default value is used.

[0014] In a preferred embodiment, the total cache capacity is converted into a total backtracking budget available for backtracking interception. The total backtracking budget is determined based on the ring cache capacity and the amount of video data written per unit time. The depth allocation module uses a piecewise linear saturation mapping function to map the unilateral compression index to the allocation ratio corresponding to the backtracking depth of the driver status video. The backtracking depth of the driver status video is determined based on the total backtracking budget and the allocation ratio. The backtracking depth of the forward environment video is then determined based on the remaining backtracking budget. The sum of the backtracking depths of the driver status video and the forward environment video does not exceed the total backtracking budget.

[0015] In a preferred embodiment, the information uploaded to the platform includes a linkage evidence package. The system also includes a video capture module, which is used to capture driver state video segments and forward environment video segments from the local cache based on the anchor point time of the actual sideslip event, the backtracking depth of the driver state video, and the backtracking depth of the forward environment video. The information uploaded to the platform includes a linkage evidence package, which at least includes the driver state video segment, the forward environment video segment, the anchor point time of the actual sideslip event, and the location coordinate information corresponding to the anchor point time of the actual sideslip event, and records the corresponding position of the anchor point time of the actual sideslip event in the two video segments and the identifiers of the two video segments.

[0016] The effects and advantages of the vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS provided by this invention are as follows: The present invention provides a vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS. By performing continuous difference entry gating on the aligned BeiDou and GPS heading change difference sequences and generating entry trigger times to limit the effective working time window for orderly processing, the trigger criterion under unilateral continuous obstruction conditions is transformed from a single peak observation to an executable interval-based entry criterion, thereby reducing write pollution caused by dense pseudo-trigger writes. After entry, a jump activity index is generated and combined with the lateral acceleration integral trend index to perform true and false sideslip cross-judgment, outputting a pseudo-trigger interception signal or the anchor time of the real sideslip event. This ensures that real wall-hugging sideslip can still be stably anchored under continuous difference background, reducing missed reports and records caused by simple frequency limiting. Furthermore, the single-sided pressure index is updated by the false trigger interception signal during the statistical period, and the driver backtracking depth and forward backtracking depth are allocated and locked under the total backtracking budget constraint. This allows backtracking resources to be adjusted according to pollution pressure and the occurrence of event anchors, ensuring backtracking coverage depth when real events occur and reducing the uploading of invalid evidence packets under the conditions of fixed cache capacity and limited uplink bandwidth. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the system structure of the present invention; Figure 2 This is a schematic diagram comparing the timing of determining true and false sideslip and the asymmetric backtracking depth allocation under unilateral occlusion conditions. Detailed Implementation

[0018] 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.

[0019] This invention provides a vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS. Vehicle-mounted video linkage monitoring is commonly used to continuously collect forward environmental video streams and driver status video streams at the vehicle end and write them to a local circular buffer. This allows for the backtracking and extraction of dual-channel video clips as evidence when an event is triggered, which are then uploaded to a platform for linked playback. Existing solutions typically generate event triggers based on changes in BeiDou and GPS heading, coupled with trigger frequency limiting and fixed backtracking depth configurations. Their default premise is that abnormal heading changes are sufficient to indicate a genuine deviation, and that the frequency limiting and fixed backtracking configurations can operate stably under fixed buffer resources and uplink bandwidth limitations.

[0020] However, under continuous unilateral obstruction, GPS heading changes are prone to short-term spikes, while BeiDou heading changes are relatively more stable, resulting in a persistent difference between the two after time alignment. In this situation, without executable criteria and orderly processing for persistent differences, the system may misinterpret the spikes superimposed on the persistent differences as genuine sideslip and frequently trigger segment solidification, leading to dense pseudo-trigger writes and rapid overwriting of critical historical segments under fixed resource constraints, resulting in overwrite pollution. Conversely, simply blocking difference-based triggers to suppress overwriting may lead to missed reports and records when genuine slow, close-to-the-wall sideslip occurs, creating an engineering conflict between suppressing pseudo-trigger overwrite pollution and avoiding missed reports of genuine close-to-the-wall sideslip, and reducing the backtracking coverage depth when real events occur.

[0021] To this end, this invention uses the aligned BeiDou heading change and GPS heading change difference sequence as the entry point. It first performs continuous difference entry gating and generates entry trigger times to limit the working range for orderly processing of unilateral obstruction. Upon entry, it generates a jump activity index and combines it with a lateral acceleration integral trend index to perform true / false sideslip cross-judgment, outputting either a false trigger interception signal or the anchor time of a true sideslip event. Subsequently, within the statistical period, it updates the unilateral pressure index based on the false trigger interception signal, and performs asymmetric backtracking allocation and locking on the driver backtracking depth and forward backtracking depth under the constraint of the total backtracking budget. This supports the generation of linked evidence packages while simultaneously suppressing overwrite contamination and ensuring backtracking coverage depth.

[0022] Based on the above design, this invention constructs a vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS, consisting of a data acquisition module U101, a difference triggering module U102, a authenticity discrimination module U103, a depth allocation module U104, and a video capture module U105, as follows: The acquisition module U101, serving as the front-end data intake point of the system link, generates a unified timestamp based on the received heterogeneous vehicle perception data, completing multi-source data synchronization and cache preparation. The heterogeneous vehicle perception data is limited to the forward environment video stream X101, driver status video stream X102, BeiDou positioning output X103, GPS positioning output X104, and inertial measurement unit output X105, along with their unified timestamped sample records. This module, on one hand, writes the dual-channel video streams to a local circular cache B101 with a unified timestamp to support subsequent backtracking and retrieval; on the other hand, it calculates and generates a heading change and difference result set R101 and writes it to a single-sided occlusion handling state container C101, thus providing consistent underlying feature input for subsequent gating and cross-judgment.

[0023] The difference trigger module U102 reads the heading change and difference result set R101 from the single-sided occlusion handling status container C101. Based on the aligned difference sequence, it determines whether it meets the continuous judgment duration condition, thereby triggering the orderly processing of single-sided occlusion and generating a continuous difference that enters the gating result set R102 and is written to C101. R102 is used to record the entry gating status, the entry trigger time, and the start and end times of the effective working time window, so that the subsequent authenticity discrimination module can perform cross-judgment under the same window caliber.

[0024] The authenticity discrimination module U103, as the core analysis and judgment node of the system link, is used to read R101 and R102 in C101 after the gating is established. Within the effective working time window, it generates a jump activity index and cross-judges it with the lateral acceleration integral trend index, generating a true / false sideslip cross-judgment result set R103, which is then written into C101. R103 is used to record the anchor point time of false trigger interception signals or real sideslip events, providing a judgment basis for subsequent backtracking allocation and evidence package generation.

[0025] The depth allocation module U104 reads R103 from C101, updates the unilateral compression index based on the pseudo-trigger interception signal during the statistical period, and locks the driver's backtracking depth and forward backtracking depth under the total backtracking budget constraint after the actual sideslip event anchor point is generated. It then generates an asymmetric backtracking allocation result set R104 and writes it to C101. R104 records the unilateral compression index, total backtracking budget, locked driver's backtracking depth and forward backtracking depth, and locking marker, providing consistent capture boundary parameters for the video capture module.

[0026] The video capture module U105, acting as the end-point execution and output node of the system link, reads R103 and R104 from C101 and extracts driver status video segments and forward environment video segments from the local circular cache B101 according to the locked backtracking depth. The module prioritizes extracting anchor point time and location coordinate information from the GPS positioning output X104. If the GPS positioning output X104 is missing or invalid, it extracts location coordinate information with the same timestamp from the Beidou positioning output X103, encapsulates it to generate a linkage evidence package, and uploads it to the platform for linkage playback.

[0027] The implementation process and operating effects of the system of the present invention will be described in detail below with reference to specific embodiments. It should be understood that the embodiments are only used to illustrate the technical solution of the present invention, and not to limit it. Without changing the essence of the invention, the relevant steps, parameters and module divisions can be appropriately adjusted.

[0028] To facilitate implementation, this system introduces a preset parameter storage area to load and store preset parameters and rule entries, and provides a call interface for querying by entry name to the acquisition module U101, difference trigger module U102, authenticity discrimination module U103, depth allocation module U104, and video capture module U105. The entries include at least the following: BeiDou heading change calculation caliber, GPS heading change calculation caliber, time alignment caliber, difference sequence definition caliber, continuous judgment duration condition caliber, difference threshold, velocity threshold, jump threshold, activity threshold, lateral acceleration integral trend index caliber, lateral acceleration zero bias correction caliber, low dynamic condition judgment caliber, sideslip trend threshold name, trend window name, unilateral compression index conversion caliber, statistical period name, total backtracking budget conversion caliber, piecewise linear saturated mapping function segment point name, authenticity sideslip cross-judgment output priority caliber, ordered processing effective working time window limitation caliber, and anchor point time segment position mapping caliber. After loading an entry, the module generates a configuration version identifier and carries this identifier in the interface response, ensuring that all modules run under the same configuration version. For threshold-type entries, the module returns both the threshold entry and its corresponding threshold value in the query results. The threshold value is used for comparison and judgment by subsequent modules, and the threshold value is not fixed locally within the module. When a requested entry is missing or unavailable, the module returns a missing flag. The calling module then uses the default entry value or the most recently available entry value from the previous version to perform a rollback. If the entry is still unavailable, the corresponding judgment branch is set to invalid or the corresponding sample is marked as not participating, and the rollback flag is registered in the result set for subsequent modules to identify.

[0029] This system uses a single-sided occlusion handling state container C101 to store the outputs R101 to R104 from each module. Each module writes or updates the corresponding result set to C101 using a unified timestamp as the index key, and reads the required result set from C101 according to the unified timestamp during subsequent processing, enabling gating, judgment, allocation, and backtracking to be executed collaboratively under the same index caliber. For R101, container C101 stores the data according to the unified timestamp of the sample and supports reading sample records within the required time range in the order of the unified timestamp. For R102 to R104, container C101 stores the data according to the unified timestamp of the event, where R102 uses the unified timestamp corresponding to the entry trigger time as the index key, R103 uses the anchor time of the actual sideslip event or the sample time corresponding to the spurious trigger interception signal as the index key, and R104 uses the unified timestamp corresponding to the anchor time of the actual sideslip event as the index key.

[0030] In an optional embodiment, the acquisition module U101 performs the following processes: data intake and unified time mapping, speed gating and dual-mode heading difference calculation, lateral acceleration integral trend index generation, and result set generation and writing. During data intake and unified time mapping, the module continuously acquires the forward environment video stream X101 and the driver status video stream X102, and simultaneously acquires the BeiDou positioning output X103, GPS positioning output X104, and inertial measurement unit output X105. The BeiDou positioning output X103 includes at least heading value and position coordinates; the GPS positioning output X104 includes at least heading value, position coordinates, and vehicle speed information; and the inertial measurement unit output X105 includes at least lateral acceleration sampling. The module generates a unified timestamp for the aforementioned video stream and sensor sampling based on the unified clock source configured at the system's underlying level. It then writes X101 and X102 into the local circular buffer B101's cache storage space based on this unified timestamp. Simultaneously, it continuously maintains the write pointer and the mapping relationship between the unified timestamp and the cache location, enabling subsequent modules to backtrack and locate the start and end range of the corresponding segment based on the actual side-slip event anchor point. The module obtains the values ​​and caliber constraints of the required entries by calling the query interface by entry name in the preset parameter storage area, and also obtains the configuration version identifier.

[0031] In speed gating and dual-mode heading difference calculation, the module retrieves the speed threshold and its corresponding value from the preset parameter storage area, and compares the vehicle speed information in the GPS positioning output X104 with the speed threshold value. When the vehicle speed is not lower than the speed threshold value, the module allows heading calculation. When the vehicle speed is lower than the speed threshold value, the module registers the sample at that moment as not participating in the marking, for subsequent gating and decision-making to skip that moment sample.

[0032] In the release state, the module calculates the BeiDou and GPS heading changes by differing the heading values ​​at adjacent sampling times, based on both the BeiDou and GPS heading change calculation methods. To avoid falsely large differences when the heading values ​​cross zero degrees or a full circle, the module performs angle normalization on the difference results, mapping the differences to a preset continuous angle interval. This preset continuous angle interval is defined or returned by the BeiDou and GPS heading change calculation methods. Subsequently, the module performs time alignment on the two heading change sequences according to the time alignment method. As one possible implementation, the module constructs a common timeline based on the unified timestamp of the GPS heading change sequence and performs nearest neighbor matching alignment on the BeiDou heading change sequence, while limiting the maximum alignment tolerance according to the time alignment method. Sample pairs exceeding the maximum alignment tolerance are registered as alignment missing markers and skipped from the difference calculation. The module further generates an aligned difference sequence based on the definition of the difference sequence. The difference sequence is generated by subtracting the aligned BeiDou heading change from the aligned GPS heading change and retaining the sign. The absolute value of the difference sequence is used for the threshold determination of subsequent gating. The threshold value used for the threshold determination is returned by the preset parameter storage area based on the difference threshold.

[0033] In generating the lateral acceleration integral trend index, the module monitors the continuous lateral acceleration sampling data (X105) output by the inertial measurement unit in real time according to the low-dynamic condition judgment criteria. The low-dynamic condition judgment criteria can at least include that the vehicle speed change amplitude does not exceed a preset upper limit and the lateral acceleration variance does not exceed a preset upper limit. When the low-dynamic conditions are met, the lateral acceleration zero-bias correction caliber can be obtained by using an exponential sliding update method, weighting the current zero-bias estimate by the previous estimate and the current lateral acceleration sample according to a preset smoothing coefficient. Subsequently, according to the lateral acceleration integral trend index caliber, the module performs cumulative integration calculation on the zero-bias corrected lateral acceleration at sampling intervals within the sliding integration window, and outputs the window integral quantity as the lateral acceleration integral trend index, where the length of the sliding integration window is determined by the lateral acceleration integral trend index caliber.

[0034] In the generation and writing of the result set, the module encapsulates the BeiDou heading change sequence, GPS heading change sequence, aligned difference sequence, lateral acceleration integral trend index, non-participation marker, alignment missing marker, and configuration version identifier to generate a heading change and difference result set R101. Then, it writes R101 into the single-sided occlusion handling status container C101 with a unified timestamp as the index for subsequent modules to read.

[0035] In an optional embodiment, the implementation process of the difference trigger module U102 includes calling the judgment parameter and reading the difference sequence, judging the continuous over-limit, determining the trigger time backtracking, and generating and writing the gating result set.

[0036] In the parameter call and difference sequence reading, the module obtains the difference threshold and its corresponding value by calling the query interface by item name in the preset parameter storage area, and obtains the continuous judgment duration value limited by the continuous judgment duration condition. The module reads the heading change and difference result set R101 from the single-sided occlusion handling state container C101 according to a unified timestamp and extracts the aligned difference sequence, while skipping samples that do not participate in the labeling and those with missing alignment labels.

[0037] In the continuous violation determination process, the module takes the absolute value of each valid difference sample and compares it with the difference threshold. The violation duration is accumulated using the uniform timestamp difference of adjacent valid samples. If a non-violation state is encountered, the accumulated value is reset. When the violation state continues for a continuous determination duration, the module determines that gating has been established and triggers unilateral occlusion ordered processing. As a default strategy, when a sample that is not involved in labeling or corresponds to a missing label is encountered, the module considers that moment as non-violation and interrupts continuous accumulation.

[0038] In the process of determining the trigger time, when the entry gate is established, the module backtracks along the time axis in the historical direction to locate the continuous over-limit interval corresponding to this judgment, and determines the unified timestamp of the first valid sample at the beginning of the interval as the entry trigger time.

[0039] In the generation and writing of the gating result set, the module generates a continuous difference value and enters it into the gating result set R102, along with the entry gating status, the entry trigger time, and the start and end times of the effective working time window determined according to the caliber of the ordered processing effective working time window. The start time of the effective working time window is taken as the entry trigger time, and the end time of the effective working time window is taken as the entry trigger time plus the window duration value, which is returned by the caliber of the ordered processing effective working time window. The module uses the unified timestamp corresponding to the entry trigger time as an index to write R102 into the single-sided occlusion handling status container C101 for subsequent reading by the module.

[0040] In an optional embodiment, the implementation process of the authenticity discrimination module U103 includes reading the effective working time window, generating the jump activity index, performing the authenticity side-slip cross judgment, and generating and writing the judgment result set.

[0041] During the effective working time window reading, the module uses the unified timestamp corresponding to the entry trigger time as an index to read the continuous difference from the single-sided occlusion handling state container C101 and enter it into the gating result set R102. It confirms that the entry gating is established, obtains the entry trigger time and the start and end times of the effective working time window, and performs subsequent cross-judgment calculations within the effective working time window.

[0042] In generating the jump activity index, the module reads the heading change and difference result set R101 from C101 and extracts the GPS heading change sequence, skipping samples that are not included in the labeling or have missing alignment labels. The module compares the absolute value of the GPS heading change with the jump threshold value one by one, and counts the number of samples exceeding the jump threshold value within the length of the sliding time window configured and returned by the preset parameter storage area, thus generating the jump activity index.

[0043] In the true / false sideslip crossover determination process, the module extracts the lateral acceleration integral trend index from R101 and skips samples corresponding to those not included in the labeling and those with missing alignment labels. The module determines whether the jump activity index is not less than the activity threshold value to confirm the jump condition. Based on the sideslip trend threshold name, it retrieves the sideslip trend threshold value from the preset parameter storage area and determines whether the lateral acceleration integral trend index is not less than the sideslip trend threshold value within the window corresponding to the trend window name to confirm the sideslip trend condition. Based on the output priority criteria for true / false sideslip crossover determination, the module generates a false trigger interception signal when the jump condition is met but the sideslip trend condition is not met; it generates the true sideslip event anchor time when the sideslip trend condition is met; and it generates the true sideslip event anchor time when both conditions are met simultaneously. When the sideslip trend condition is met for the first time, the module determines the unified timestamp corresponding to that sample as the true sideslip event anchor time.

[0044] In the generation and writing of the judgment result set, the module generates a true / false sideslip cross judgment result set R103 and writes it to the anchor point time of the false trigger interception signal or the true sideslip event, and only writes one of them at the same index time. The module writes R103 to the single-sided occlusion handling state container C101 with the unified timestamp corresponding to the sample time of the false trigger interception signal or the anchor point time of the true sideslip event as the index.

[0045] In an optional embodiment, the implementation process of the depth allocation module U104 includes preset parameter extraction, unilateral compression index sliding statistics and state update, total backtracking budget conversion and asymmetric backtracking depth mapping, and allocation result set generation and writing.

[0046] In the preset parameter extraction, the module obtains the statistical period duration value corresponding to the statistical period name, the conversion caliber of the unilateral compression index, the conversion caliber of the total retrospective budget, the segment point name of the piecewise linear saturated mapping function, the minimum retrospective lower limit and its value, and the default value of the unilateral compression index and its value from the preset parameter storage area.

[0047] The module reads the true / false sideslip crossover determination result set R103 from the single-sided occlusion handling status container C101 according to a unified timestamp. When a false trigger interception signal is read, the module only updates the single-sided compression index and does not trigger the backtracking allocation calculation.

[0048] In the sliding statistics and status update of the unilateral compression index, the module accumulates the number of false trigger interception signals within a sliding statistical window defined by the statistical period name, based on the conversion caliber of the unilateral compression index. The number of triggers is then converted into the unilateral compression index by the duration of the statistical period. The sliding statistical window is updated on a rolling basis with a uniform timestamp, and the window length is equal to the duration of the statistical period. When no new false trigger interception signal is generated, the module retains the most recently updated value; if no most recently updated value exists, it is initialized with the default value of the unilateral compression index.

[0049] In the total backtracking budget calculation and asymmetric backtracking depth mapping, when the actual sideslip event anchor point time is read from R103, the module determines the total backtracking budget based on the total backtracking budget calculation caliber, combined with the ring buffer capacity of the local ring buffer B101 and the unit time video data write volume returned by the total backtracking budget calculation caliber. The unit time video data write volume is returned by the preset parameter storage area by default. The module determines the mapping interval based on the segment point name of the piecewise linear saturation mapping function, mapping the unilateral compression index to the driver's backtracking depth allocation ratio. The allocation ratio ranges from 0 to 1 and enters the saturation interval at the segment point. The piecewise linear saturation mapping function calculates the allocation ratio by linear interpolation between adjacent segment points, taking the lower limit ratio below the minimum segment point and the upper limit ratio above the maximum segment point. The module determines the driver's backtracking depth based on the total backtracking budget and the allocation ratio, and determines the forward backtracking depth by deducting the driver's backtracking depth from the total backtracking budget.

[0050] In the generation and writing of the allocation result set, the module compares the driver's backtracking depth and forward backtracking depth with the minimum backtracking lower limit value, respectively. When the forward backtracking depth is lower than the minimum backtracking lower limit value, the forward backtracking depth is adjusted to the minimum backtracking lower limit value and the driver's backtracking depth is deducted accordingly, so that the depths of both paths are not less than the minimum backtracking lower limit value and the sum of the two does not exceed the total backtracking budget. After completing the fallback calibration, the module locks the depths of both paths until the current backtracking interception is completed. Subsequently, the module generates an asymmetric backtracking allocation result set R104 and writes the unilateral compression index, total backtracking budget, locked driver's backtracking depth and forward backtracking depth, and locking flag. Using the unified timestamp corresponding to the anchor point time of the actual sideslip event as an index, R104 is written into the unilateral occlusion handling status container C101.

[0051] In an optional embodiment, the implementation process of the video capture module U105 includes capture context reading, dual-channel video backtracking capture, position coordinate alignment extraction, and linkage evidence package encapsulation and uploading.

[0052] During context retrieval, the module reads the true / false sideslip cross-judgment result set R103 from the single-sided occlusion handling state container C101 and obtains the true sideslip event anchor point time. Using the unified timestamp corresponding to this true sideslip event anchor point time as an index, it reads the asymmetric backtracking allocation result set R104 from C101 to obtain the locked driver backtracking depth and forward backtracking depth. When R103 does not provide the true sideslip event anchor point time and only provides a false trigger interception signal, the module does not generate a linkage evidence package and does not trigger backtracking interception.

[0053] In dual-channel video backtracking and interception, the module determines the start and end times of interception based on the actual sideslip event anchor point time and the locked driver backtracking depth and forward backtracking depth. The interception start time is the actual sideslip event anchor point time minus the driver backtracking depth, and the interception end time is the actual sideslip event anchor point time plus the forward backtracking depth. Based on the write pointer maintained in the local circular cache B101 and the mapping relationship between unified timestamps and cache locations, the module locates the corresponding frame sequence by timestamp index and exports temporally continuous driver state video segments and forward environment video segments.

[0054] In the location coordinate alignment extraction, the module prioritizes reading the GPS positioning output X104 and retrieves the location coordinate information aligned with the actual side-slip event anchor point time based on the unified timestamp. When the GPS positioning output X104 is missing coordinates at the corresponding timestamp or is marked as invalid, the module extracts the location coordinate information with the same timestamp from the Beidou positioning output X103. The determination of invalid coordinates is given by the validity mark built into the positioning output or the validity determination criterion configured in the preset parameter storage area. If no validity mark is given, it is treated as missing coordinates by default.

[0055] In the encapsulation and uploading of the linked evidence package, the module obtains the anchor point time segment position mapping caliber through the preset parameter storage area. The two video segment identifiers are generated by combining the video channel identifier and the start and end times of the capture, used to identify the video channel and segment range. The video channel identifier is used to distinguish the video channels corresponding to the forward environment video stream X101 and the driver status video stream X102. The video channel identifier is determined by the acquisition module U101 when writing to the local circular buffer B101 and read by the video capture module U105. The corresponding position of the anchor point time in the video segment is recorded as the time offset of the anchor point time relative to the segment's start time. If necessary, it can be converted from the frame rate to the frame number. The frame rate is taken from the video stream metadata or returned from the preset parameter storage area. The anchor point time segment position mapping caliber is used at least to limit the recording accuracy and conversion rules of the time offset; the default accuracy is consistent with the unified timestamp resolution. The module writes and encapsulates the driver's status video clip, forward environment video clip, the anchor point time of the actual sideslip event, position coordinate information, the identifiers of the two video clips, and the corresponding position of the anchor point time to generate a linkage evidence package, and uploads the linkage evidence package to the platform through the vehicle network uplink.

[0056] To illustrate that under the condition that continuous unilateral obstruction causes a persistent difference between the BeiDou heading change and the GPS heading change, the system can distinguish between false triggering and genuine wall-hugging sideslip after the gate is established, and perform an asymmetric allocation of the driver's backtracking depth and forward backtracking depth under a fixed total backtracking budget constraint, the following is given: Figure 2 The timing comparison diagram is shown below. Figure 2 In this context, A represents a pseudo-trigger interception scenario. Figure 2 In the diagram, B represents the actual sideslip and asymmetric distribution scenario.

[0057] Figure 2 middle, The trigger time indicates that the continuous difference has entered the gating process and the unilateral occlusion is processed in an orderly manner. The activity threshold is used to determine the jump condition, and the sideslip trend threshold is used to determine the sideslip trend condition. For the sample time corresponding to the false trigger interception signal, The anchor point for the actual sideslip event is shown. The curves in the figure are schematic normalized displays, and the threshold values ​​are mapped to the same normalization scale during the display. They are only used to illustrate the timing logic and judgment relationship. Figure 2 The asymmetric backtracking depth diagram below illustrates that, under the constraint of a fixed total backtracking budget, the driver's backtracking depth and forward backtracking depth are asymmetrically allocated, and the sum of the two segments equals the total backtracking budget. Figure 2 This is used to explain the cross-determination and asymmetric allocation logic after the entry gate is established. The entry gate determination is based on the alignment difference sequence in the heading change and difference result set R101 and the continuous difference entry gate result set R102. The relevant calculation methods have been explained in the aforementioned implementation.

[0058] like Figure 2 As shown in A, in Subsequently, a short-lived spike occurred in the GPS heading change, causing the jump activity index to reach the activity threshold within the slip time window. However, the lateral acceleration integral trend index did not reach the sideslip trend threshold. Therefore, the system determined that the jump condition was met and the sideslip trend condition was not met. A false trigger interception signal is generated, and the trigger of this backtracking interception is blocked, thereby avoiding invalid interception and evidence contamination caused by multipath interference.

[0059] like Figure 2 As shown in B, in Subsequently, when the lateral acceleration integral trend indicator first reaches the sideslip trend threshold value within the trend window, the system generates the actual sideslip event anchor point moment. Subsequently, under the fixed constraint of the total backtracking budget, the system determines the allocation ratio of driver backtracking depth and forward backtracking depth based on the unilateral pressure index obtained from the false trigger interception signal update within the statistical period. This results in a relative increase in driver backtracking depth and a relative decrease in forward backtracking depth. Based on this, the system backtracks and extracts driver status video clips and forward environment video clips from the local circular cache B101, encapsulates them to generate a linkage evidence package, and uploads it to the platform to support linkage playback.

[0060] 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.

[0061] 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.

[0062] 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.

[0063] 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.

[0064] 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 multi-device real-time positioning and video linkage system based on BeiDou and GPS, characterized in that, include: The acquisition module is used to acquire forward environment video, driver status video, Beidou heading changes, GPS heading changes, and vehicle inertial measurement unit output information, and write them to the local cache. The difference trigger module is used to enter the authenticity judgment module when the difference between the Beidou heading change and the GPS heading change meets the continuous judgment duration condition. The authenticity discrimination module is used to combine the jump state of GPS heading change with the output information of the inertial measurement unit to distinguish between false triggers and real sideslips. When it is distinguished as a false trigger, it outputs a false trigger interception signal; when it is distinguished as a real sideslip, it outputs the anchor point time of the real sideslip event. The depth allocation module is used to determine the backtracking depth of the driver status video based on the pseudo-trigger interception signals within the statistical period, under the constraint of the total backtracking budget, and then determine the backtracking depth of the forward environment video based on the remaining backtracking budget.

2. The vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS as described in claim 1, characterized in that, The acquisition module generates a unified timestamp for the forward environment video, driver status video stream and sensor samples based on a unified clock source, and performs writing and time alignment based on the unified timestamp. Among them, vehicle sensor samples include at least Beidou heading change, GPS heading change and inertial measurement unit output information.

3. The vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS according to claim 2, characterized in that, The local circular cache is a circular storage structure indexed by timestamps, which at least maintains write pointers and the mapping relationship between timestamps and cache locations.

4. The vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS according to claim 2, characterized in that, The BeiDou heading change and GPS heading change are obtained from the difference of heading values ​​at adjacent sampling times. When the vehicle speed is lower than the preset speed threshold, the BeiDou heading change and GPS heading change will not participate in the subsequent judgment. The continuous judgment duration condition includes that the absolute value of the difference between the aligned BeiDou heading change and GPS heading change is not less than the preset difference threshold and the continuous judgment duration is reached.

5. The vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS according to claim 4, characterized in that, The output information of the inertial measurement unit includes at least lateral acceleration information. The sideslip trend corresponding to the lateral acceleration information can be characterized by the lateral acceleration integral trend index. The lateral acceleration integral trend index is the window integral quantity obtained by integrating the lateral acceleration after zero bias correction within the sliding time window, and the lateral acceleration zero bias estimate is updated when the preset low dynamic conditions are met. The sideslip trend conditions include that the lateral acceleration integral trend index is not less than the preset sideslip trend threshold within the preset trend window.

6. The vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS according to claim 5, characterized in that, The jump status of GPS heading changes can be characterized by the jump activity index; the jump activity index is the number of samples in which the absolute value of the GPS heading change exceeds a preset jump threshold within a preset sliding time window, and the jump condition includes that the jump activity index is not less than the preset activity threshold.

7. The vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS according to claim 6, characterized in that, The cross-judgment of the authenticity discrimination module satisfies the following rules: when the jump condition is met and the sideslip trend condition is not met, a false trigger interception signal is generated; when the sideslip trend condition is met, the true sideslip event anchor point time is generated; and when both the jump condition and the sideslip trend condition are met, the true sideslip event anchor point time is generated.

8. The vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS according to claim 7, characterized in that, The false trigger interception signal within the statistical period is used to characterize the degree of change in the driver's status video review demand. It can be represented by the one-sided pressure index in the calculation. The one-sided pressure index is calculated by converting the number of false trigger interception signals within the preset statistical period with the preset statistical period duration. The one-sided pressure index is updated by using a sliding statistical window counting method. When no false trigger interception signal is generated, the one-sided pressure index is kept at the most recent updated value. When there is no most recent updated value, the preset default value is used.

9. The vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS according to claim 8, characterized in that, The total backtracking budget is determined based on the ring buffer capacity and the amount of video data written per unit time. The depth allocation module uses a piecewise linear saturation mapping function to map the unilateral compression index to the allocation ratio corresponding to the backtracking depth of the driver status video. The backtracking depth of the driver status video is determined based on the total backtracking budget and the allocation ratio. The backtracking depth of the forward environment video is then determined based on the remaining backtracking budget. The sum of the backtracking depths of the driver status video and the forward environment video does not exceed the total backtracking budget.

10. The vehicle-mounted multi-device real-time positioning and video linkage system based on BeiDou and GPS according to claim 9, characterized in that, The system also includes a video capture module, which is used to capture driver state video clips and forward environment video clips from the local cache based on the anchor point time of the actual sideslip event, the backtracking depth of the driver state video, and the backtracking depth of the forward environment video. The information uploaded to the platform includes a linkage evidence package, which includes at least the driver state video clip, the forward environment video clip, the anchor point time of the actual sideslip event, and the location coordinate information corresponding to the anchor point time of the actual sideslip event. It also records the corresponding position of the anchor point time of the actual sideslip event in the two video clips and the identifiers of the two video clips.

Images