Highway management system based on event response

Abstract

The invention relates to the field of intelligent road management, in particular to a road management system based on event response, which is characterized in that a multi-source sensing module is arranged to acquire an original code stream of an acquisition terminal and extract coding domain characteristics to construct a time domain change curve, and the time domain change curve is segmented through an association processing module; the method comprises the following steps: obtaining curve segments, classifying the obtained curve segments, constructing an evidence association relationship between the curve segments, carrying out evidence association analysis by an analysis module according to the categories of the curve segments, setting evidence association tags for the curve segments, and carrying out decoding analysis on original code streams in each time domain interval by an intelligent processing module according to the evidence association tags. The response module is used for performing exception handling. According to the method, decoding analysis is carried out on the curve segment provided with the evidence association tag adaptively, so that the verification number of decoding frames is reduced, the reliability and the recognition efficiency of anomaly recognition are ensured, the anomaly information is pushed to the response end and the roadside display board in time, and the emergency disposal efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent highway management, and more particularly to an event-response-based highway management system. Background Technology

[0002] Achieving accurate monitoring and efficient emergency response to traffic accidents has become a core issue in ensuring the safety of highway operations. Currently, with the rapid development of artificial intelligence, the Internet of Things, big data and communication technologies, large-scale model and intelligent agent technologies, with their core advantages of autonomous perception, autonomous decision-making and collaborative interaction, provide a new technical path for solving the collaborative linkage problem in highway emergency response.

[0003] For example, Chinese Patent Publication No. CN119942802A discloses a method and device for coordinated management and control of emergency events on highways. The method includes: responding to the acquisition of lane images from different sections of the highway; identifying the images to obtain at least lane traffic condition information; determining, based on the traffic conditions, that the vehicle speed in a lane is below a first threshold, or the vehicle density exceeds a second threshold, or the distance between vehicles in multiple lanes is less than a third threshold; determining that traffic intervention is required for the corresponding highway section; identifying the cause of congestion based on the acquired images of the section to be intervened; determining the intervention method based on the location information and cause of congestion of the section to be intervened; and notifying local authorities of each intervention method according to the intervention method for the section to be intervened, with each authority then intervening in the traffic flow on the highway according to the intervention method. This application ensures the efficiency of highway traffic while reducing the accident rate. However, the following problems still exist in the existing technology. In the existing highway monitoring system, the raw code streams generated by massive acquisition terminals need to be decoded and analyzed across the entire domain, resulting in a waste of computing resources, high computation time, and poor real-time performance. In addition, the regularity of traffic flow in the time domain under a fixed viewpoint of a single terminal is not considered, and there is a lack of mutual verification between adjacent time domains, which reduces the efficiency of abnormal event identification. Summary of the Invention

[0004] To address this, the present invention provides an event-response-based highway management system to overcome the problems in the existing technology, which suffer from high computational overhead and poor real-time performance due to the full-domain decoding of massive amounts of raw code streams from terminals in the existing highway monitoring system; and the lack of consideration for the temporal variation patterns of traffic flow from individual terminals and the absence of mutual verification between adjacent temporal domains, resulting in low efficiency in identifying abnormal events.

[0005] To achieve the above objectives, the present invention provides an event-response-based highway management system, comprising: The multi-source sensing module is used to acquire the raw bitstreams collected by several acquisition terminals and extract coding domain features based on the raw bitstreams to construct time-domain variation curves. The correlation processing module is used to segment the time-domain change curves based on curve features, classify the obtained curve segments, and group and associate them according to the time sequence of the same type of curve segments to build corroborative correlation relationships between each curve segment. The analysis module is used to identify curve segments with corroborating relationships, and performs corroborating relationship analysis based on the category corresponding to the curve segment, including: The inflection points in the calibrated curve segments are determined, and several micro-observation time domains are determined based on each inflection point to extract the curve change characteristics within each micro-observation time domain. Based on the differences in curve change characteristics between curve segments, corroborating association labels are set for the curve segments. Each curve segment is extracted and placed in the same coordinate system. Based on the overlap of the curve segments, corroborating association labels are set for the curve segments. The intelligent processing module is used to determine the time domain interval corresponding to each curve segment, and to decode and analyze the original bitstream within each time domain interval based on supporting correlation tags, including: Based on the category of the curve segment, the original bitstream is extracted to obtain the decoded frame and determine the verification label. The verification label is then assigned to all the original bitstreams within the time domain interval. The original bitstream is time-series decoded to obtain decoded frames, and the verification tags of the original bitstream are determined sequentially based on the decoded frames.

[0006] Furthermore, the multi-source sensing module extracts coding domain features from the original bitstream to construct a time-domain variation curve, including: The proportion of target macroblocks and the macroblock division depth of the corresponding video frames at each time point are determined based on the original bitstream. The coding domain features are obtained by normalizing the proportion of the target macroblock and the macroblock partitioning depth, and then summing them by weight. A time-domain variation curve is constructed with the coding domain features as the vertical axis and time as the horizontal axis. The target macroblock is a non-skipped macroblock.

[0007] Furthermore, the correlation processing module is used to segment the time-domain variation curve based on curve features, including: Construct time windows and slide them on the time axis to determine the curve characteristics of each time window, including the floating frequency and the average amplitude; Verify the rate of change of the curve features corresponding to the adjacent time windows in the verification time window, lock the segmented time window based on the rate of change, and determine the first moment of the segmented time window as the segmentation moment; The time-domain variation curve is divided into several curve segments based on the segmentation time. If the rate of change of a curve feature is greater than the predetermined rate of change of the segmentation threshold, then the time window is locked as the segmentation time window.

[0008] Furthermore, the association processing module is used to classify the obtained curve segments, group and associate them according to the temporal order of the same type of curve segments, and construct corroborating association relationships between the curve segments, including: If a curve segment meets the floating condition, the curve segment is classified as a feature-dense type. If the curve segment does not meet the floating condition, the curve segment is classified as a non-feature dense type. Identify temporally adjacent curve segments of the same type, verify the maximum time interval between curve segments of the same type, and if the maximum time interval is less than a predetermined interval threshold, determine that there is a corroborating correlation between curve segments of the same type. The floating condition is that the floating frequency of the curve segment is greater than a predetermined floating frequency threshold.

[0009] Furthermore, the analysis module performs corroborative correlation analysis based on the category corresponding to the curve segment, wherein, For curve segments with dense features, inflection points in the calibrated curve segments are selected, and several micro-observation time domains are determined based on each inflection point to extract curve change features within each micro-observation time domain. Based on the differences in curve change features between curve segments, corroborating association labels are set for the curve segments. For curve segments that are not feature-dense, select each curve segment and place it in the same coordinate system. Based on the overlap of the curve segments, set corroborating association labels for the curve segments.

[0010] Furthermore, the analysis module extracts the curve change features within each micro-observation time domain, and sets corroborating association labels for the curve segments based on the differences in curve change features between them. Determine the intermediate time between each inflection point and other adjacent inflection points, and define the time interval formed by two intermediate times as the micro-observation time domain; Extract the curve change characteristics of the curve segment within the micro-observation time domain, including the mean change amplitude and the mean change slope; Compare the curve change characteristics between curve segments to determine the difference ratio of curve change characteristics; Specifically, if the curve change feature difference ratio is greater than or equal to a predetermined difference ratio threshold, no supporting association label is set for the curve segment; if the curve change feature difference ratio is less than the predetermined difference ratio threshold, then supporting association label is set for the curve segment.

[0011] Furthermore, the analysis module extracts each curve segment and places them in the same coordinate system. Based on the overlap of the curve segments, it sets supporting association labels for the curve segments, including... The remaining curve segments are trimmed according to the time length corresponding to the smallest curve segment so that each curve segment corresponds to the same time length. Place all the trimmed curve segments in the same coordinate system and calculate the degree of overlap between the curve segments; If the overlap is greater than or equal to a predetermined overlap threshold, a supporting association label is set; if the overlap is less than the predetermined overlap threshold, no supporting association label is set.

[0012] Furthermore, the intelligent processing module extracts the original bitstream and obtains the decoding frame according to the category of the curve segment for the curve segment set with the corroborating association label, and determines the verification label, and assigns the verification label to all the original bitstreams in the time domain interval. For curve segments without supporting labels, time-series decoding is performed on the original bitstream to obtain decoded frames, and the verification labels of the original bitstream are determined sequentially based on the decoded frames.

[0013] Furthermore, the intelligent processing module determines the verification tag corresponding to the original bitstream based on the decoded frame, including: Determine whether the predetermined target exists in the decoded frame; If a predetermined target exists, set a verification tag for the original bitstream corresponding to the decoded frame; If no predetermined target exists, do not set a verification tag for the original bitstream corresponding to the decoded frame.

[0014] Furthermore, it also includes a response module that responds to the setting of a verification tag on the original bitstream, used to lock the acquisition terminal based on the original bitstream and perform anomaly handling, including... Based on the correlation between the data acquisition terminal and location information, abnormal locations are determined, and the abnormal locations and corresponding decoded frames are sent to the processing terminal. Simultaneously, the abnormal locations are pushed to the corresponding roadside display boards.

[0015] Compared with existing technologies, this invention acquires the raw bitstream from the acquisition terminal by setting up a multi-source sensing module and extracts coding domain features to construct a time-domain variation curve. A correlation processing module segments the time-domain variation curve, classifies the obtained curve segments, and establishes corroborating correlation relationships between the curve segments. An analysis module performs corroborating correlation analysis based on the category of the curve segments and sets corroborating correlation labels for the curve segments. An intelligent processing module decodes and analyzes the raw bitstream within each time-domain interval based on the corroborating correlation labels. A response module handles anomalies. This invention adaptively performs corroborating correlation analysis on the categories of curve segments with corroborating correlation relationships, classifies and compares curve segments with high homogeneity in a short period of time for corroboration, and then sets corroborating correlation labels to distinguish curve segments with potential anomalies. Subsequent decoding analysis adaptively performs decoding analysis on curve segments with set corroborating correlation labels, thereby reducing the number of decoding frame verifications and ensuring the reliability and efficiency of anomaly identification.

[0016] In particular, this invention constructs a temporal variation curve based on the extracted coding domain features from the original bitstream. The coding domain features can be directly extracted from the original bitstream without the need for post-decoding analysis of decoded frames, facilitating rapid analysis at the front end and reflecting changes in the image. Furthermore, the temporal variation curve is segmented using curve features. In practice, acquisition terminals are often single-viewpoint devices, and changes in the image are mainly caused by vehicles in the image. Under normal traffic conditions, the image in the short time domain exhibits certain regularities, and these regularities vary depending on the traffic situation. For example, when there are many vehicles, the temporal variation curve changes drastically due to vehicles entering and leaving the image. However, due to the overall driving characteristics of vehicles in the image, the changes also exhibit regularity even without anomalies. Similarly, when there is no traffic flow, the image is generally stable, but the coding domain features remain stable at different levels. Moreover, since traffic flow changes are non-transient, curve segments with shorter time intervals under anomaly-free conditions can be used for mutual verification, reliably and quickly distinguishing between anomalies. Supporting labels are set, and subsequent decoding analysis is performed adaptively on the curve segments with supporting labels to reduce the number of decoded frame verifications and ensure the reliability and efficiency of anomaly identification.

[0017] In particular, this invention adaptively uses correlation analysis based on the category of curve segments. For curve segments with dense features, the micro-observation time domain is determined based on inflection points. In reality, for curve segments with dense features, changes in the curve segments are more pronounced at the moments when vehicles enter and exit, easily forming inflection points. Since traffic flow changes are non-transient, these changes exhibit certain regularities and similarities under normal vehicle driving conditions. Therefore, a micro-observation time domain is constructed to observe the curve change characteristics within the time domain where changes are more pronounced. This invention does not employ a method that captures obvious abrupt changes; under multiple traffic flows, the time domain change curves fluctuate significantly, which would... Because the sporadic nature of images leads to lower accuracy, this invention considers curve segments with corroborating relationships in the absence of anomalies as potential sources of verification. By comparing the curve variation characteristics of these segments, mutual verification can be achieved with relatively low computational power, ensuring high accuracy and rapid setting of corroborating relationship labels. Furthermore, for curve segments that are not feature-dense and have small or no fluctuations, corroborating relationship curve segments are used to quickly set corroborating relationship labels by comparing similarity. This facilitates subsequent adaptive decoding analysis of the curve segments with corroborating relationship labels, thereby reducing the number of decoding frame verifications and ensuring the reliability and efficiency of anomaly identification.

[0018] In particular, the present invention performs decoding analysis on the original bitstream in each time domain interval based on the corroborating association label. For the curve segment with the corroborating association label, it reflects that there is no potential anomaly after the corroborating association analysis. Therefore, subsequently, after extracting and decoding a local original bitstream, the verification label is determined and the verification label is assigned to all original bitstreams in the time domain interval to ensure the accuracy of identification and significantly reduce the number of decoding frames involved in the verification. Attached Figure Description

[0019] Figure 1 A simplified structural diagram of an event-response-based highway management system according to an embodiment of the invention; Figure 2 A logical block diagram illustrating the corroborative relationships between curve segments in an embodiment of the invention; Figure 3 The following is a logical block diagram of the analysis module in the embodiment of the invention, which performs corroborative correlation analysis based on the category corresponding to the curve segment. Figure 4 This is a logic block diagram for decoding and analyzing the original bitstream in each time domain interval based on the evidence association tags in the embodiments of the invention. Detailed Implementation

[0020] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0021] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0022] It should be noted that in the description of this invention, the terms "upper," "lower," "left," "right," "inner," and "outer," etc., which indicate directional or regional relationships, are based on the directional or regional relationships shown in the accompanying drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0023] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0024] Please see Figure 1 The diagram shown is a simplified structural diagram of an event-response-based highway management system according to an embodiment of the invention. The event-response-based highway management system of this embodiment includes... The multi-source sensing module is used to acquire the raw bitstreams collected by several acquisition terminals and extract coding domain features based on the raw bitstreams to construct time-domain variation curves. The correlation processing module is used to segment the time-domain change curves based on curve features, classify the obtained curve segments, and group and associate them according to the time sequence of the same type of curve segments to build corroborative correlation relationships between each curve segment. The analysis module is used to identify curve segments with corroborating relationships, and performs corroborating relationship analysis based on the category corresponding to the curve segment, including: The inflection points in the calibrated curve segments are determined, and several micro-observation time domains are determined based on each inflection point to extract the curve change characteristics within each micro-observation time domain. Based on the differences in curve change characteristics between curve segments, corroborating association labels are set for the curve segments. Each curve segment is extracted and placed in the same coordinate system. Based on the overlap of the curve segments, corroborating association labels are set for the curve segments. The intelligent processing module is used to determine the time domain interval corresponding to each curve segment, and to decode and analyze the original bitstream within each time domain interval based on supporting correlation tags, including: Based on the category of the curve segment, the original bitstream is extracted to obtain the decoded frame and determine the verification label. The verification label is then assigned to all the original bitstreams within the time domain interval. The original bitstream is time-series decoded to obtain decoded frames, and the verification tags of the original bitstream are determined sequentially based on the decoded frames.

[0025] Specifically, there are no restrictions on the form of the acquisition terminal, which includes several video acquisition devices used to acquire images of highways.

[0026] Specifically, there are no specific limitations on the structure of the multi-source sensing module, the associated processing module, the analysis module, the intelligent processing module, and the response module. They can all be composed of logic components or combinations of logic components. Logic components include field-programmable processors, computers, or microprocessors in computers.

[0027] Specifically, the raw bitstream is the time-series encoded data after being encoded and processed by the acquisition terminal, which is then decoded by the decoder to form video frames, which will not be elaborated further.

[0028] Specifically, the multi-source sensing module extracts coding domain features from the original bitstream to construct a time-domain variation curve, including: The proportion of target macroblocks and the macroblock division depth of the corresponding video frames at each time point are determined based on the original bitstream. The coding domain features are obtained by normalizing the proportion of the target macroblock and the macroblock partitioning depth, and then summing them by weight. A time-domain variation curve is constructed with the coding domain features as the vertical axis and time as the horizontal axis. The target macroblock is a non-skipped macroblock.

[0029] It is understandable that a skipped macroblock refers to a special type of inter-frame prediction macroblock in video coding. This macroblock does not transmit any motion vector information or prediction residual data in the bitstream, but is only marked by a very low bit syntax element. During decoding, its motion vector is derived from the adjacent macroblock, and the pixel value is directly copied from the corresponding position in the reference frame. Therefore, the skipped macroblock corresponds to the static region in the video frame that has no change relative to the reference frame.

[0030] The number of non-skipped macroblocks reflects the size of the moving region in a video frame. When there is a continuous flow of traffic, the number of non-skipped macroblocks remains at a high level; while when vehicles enter or leave the frame or the traffic density changes, the number will fluctuate accordingly, thus providing a basis for event detection.

[0031] The macroblock segmentation depth mean is the arithmetic mean of the segmentation depths of all macroblocks in the current frame. The encoder adaptively segments macroblocks based on the image content: large, low-depth macroblocks are used for flat areas such as the sky and roads; while small, high-depth macroblocks are used for textured areas such as vehicle edges, license plates, and headlights. Therefore, the macroblock segmentation depth mean can effectively represent the proportion of textured areas in the image, and thus indirectly reflect the density of traffic flow.

[0032] In practice, the mean depth of macroblock partitioning and the proportion of target macroblocks are normalized to [0, 1]. When performing weighted summation, considering that the changing trends of the mean depth of macroblock partitioning and the proportion of target macroblocks are in the same direction, equal weight fusion can enhance the data representation. Therefore, in order to comprehensively consider the influence of the mean depth of macroblock partitioning and the proportion of target macroblocks, the weight of the mean depth of macroblock partitioning is set to 0.5, and the weight of the proportion of target macroblocks is set to 0.5.

[0033] Specifically, the correlation processing module is used to segment time-domain variation curves based on curve features, including: Construct time windows and slide them on the time axis to determine the curve characteristics of each time window, including the floating frequency and the average amplitude; Verify the rate of change of the curve features corresponding to the adjacent time windows in the verification time window, lock the segmented time window based on the rate of change, and determine the first moment of the segmented time window as the segmentation moment; The time-domain variation curve is divided into several curve segments based on the segmentation time. If the rate of change of a curve feature is greater than the predetermined rate of change of the segmentation threshold, then the time window is locked as the segmentation time window.

[0034] Understandably, the fluctuation frequency refers to the number of times the curve value changes in the direction of reversal per unit time, used to characterize the dynamic rhythm of traffic flow changes. Within a time window, the change in the sign of the difference between curve values ​​at adjacent moments is detected. The difference between the curve value at each moment and the previous moment is recorded. If the signs of the differences between adjacent moments are opposite, i.e., changing from rising to falling or from falling to rising, it is counted as a fluctuation event. The total number of fluctuation events within the window is counted and divided by the window duration to obtain the fluctuation frequency. The higher the value, the more frequent the traffic flow; the lower the value, the more stable the traffic flow. The average amplitude is the mean of the curve values ​​at each time point.

[0035] In practice, the length of the time window is set to the duration of 10 frames to ensure that the time window can contain multiple P frames, thereby obtaining sufficient statistical samples so that the curve features can reflect the changes in traffic flow.

[0036] In implementation, the purpose of setting the segmentation threshold change rate is to reflect changes in traffic flow. The segmentation threshold change rate is determined adaptively, using the curve characteristic change rates of the 30 time windows preceding the current moment as historical samples. The mean and standard deviation are calculated, and the mean plus twice the standard deviation is used as the segmentation threshold. When the change rate of a certain window exceeds this threshold, it is considered a significant change and locked as the segmentation time window. Based on this, the segmentation threshold can dynamically adapt, learning the changing patterns of curve characteristics within historical periods. When a new time window shows significant change, segmentation is triggered, thereby distinguishing different traffic flow conditions.

[0037] This invention constructs a temporal variation curve based on the extracted coding domain features from the original bitstream. These features can be directly extracted from the original bitstream without the need for post-decoding analysis of decoded frames, facilitating rapid analysis at the front end and reflecting changes in the image. Furthermore, the temporal variation curve is segmented using curve features. In practice, acquisition terminals often have a single viewpoint, and image changes are primarily caused by vehicles in the image. Under normal traffic conditions, the image exhibits certain regularities within a short time domain, and these regularities vary depending on the traffic situation. For example, when there are many vehicles, the temporal variation curve changes drastically due to vehicles entering and leaving the image. However, due to the overall driving characteristics of vehicles in the image, the changes also exhibit regularity even without anomalies. Similarly, in the absence of traffic, the image is generally stable, but the coding domain features remain stable at different levels. Since traffic changes are non-transient, curve segments with shorter time intervals in the absence of anomalies can be used for mutual verification, reliably and quickly distinguishing between anomalies. Supporting labels are set, and subsequent decoding analysis is adaptively performed on curve segments with these supporting labels, thereby reducing the number of decoded frame verifications and ensuring the reliability and efficiency of anomaly identification.

[0038] For details, please refer to Figure 2 As shown, Figure 2 This is a logical block diagram illustrating the construction of corroborative correlations between curve segments according to an embodiment of the invention. The correlation processing module is used to classify the obtained curve segments, group them according to the temporal order of similar curve segments, and construct corroborative correlations between curve segments, including... If a curve segment meets the floating condition, the curve segment is classified as a feature-dense type. If the curve segment does not meet the floating condition, the curve segment is classified as a non-feature dense type. Identify temporally adjacent curve segments of the same type, verify the maximum time interval between curve segments of the same type, and if the maximum time interval is less than a predetermined interval threshold, determine that there is a corroborating correlation between curve segments of the same type. The floating condition is that the floating frequency of the curve segment is greater than a predetermined floating frequency threshold.

[0039] In practice, the purpose of setting a floating frequency threshold is to reflect situations with less traffic. Based on this, several video frames with no traffic or no vehicles entering or leaving simultaneously, collected by the same acquisition terminal, are obtained as samples in advance. The corresponding original bitstream is determined, and the temporal variation curve corresponding to the coding domain features is constructed to determine the average floating frequency, so as to reflect the floating frequency level under situations with no traffic or less traffic. In practice, the average floating frequency is set as the floating frequency threshold.

[0040] The purpose of setting an interval threshold is to find similar curve segments within a shorter time to establish corroborating correlations and reduce the impact of environmental changes and traffic flow changes. Based on this, the interval threshold is selected within the range [5s, 10s], preferably 5s.

[0041] For details, please refer to Figure 3 The diagram shown is a logical block diagram of the analysis module in an embodiment of the invention, which performs corroborative correlation analysis based on the category corresponding to the curve segment. The analysis module performs corroborative correlation analysis based on the category corresponding to the curve segment. For curve segments with dense features, inflection points in the calibrated curve segments are selected, and several micro-observation time domains are determined based on each inflection point to extract curve change features within each micro-observation time domain. Based on the differences in curve change features between curve segments, corroborating association labels are set for the curve segments. For curve segments that are not feature-dense, select each curve segment and place it in the same coordinate system. Based on the overlap of the curve segments, set corroborating association labels for the curve segments.

[0042] Specifically, the analysis module extracts the curve change features within each micro-observation time domain, and sets corroborating association labels for the curve segments based on the differences in curve change features between them. Determine the intermediate time between each inflection point and other adjacent inflection points, and define the time interval formed by two intermediate times as the micro-observation time domain; Extract the curve change characteristics of the curve segment within the micro-observation time domain, including the mean change amplitude and the mean change slope; Compare the curve change characteristics between curve segments to determine the difference ratio of curve change characteristics; Specifically, if the curve change feature difference ratio is greater than or equal to a predetermined difference ratio threshold, no supporting association label is set for the curve segment; if the curve change feature difference ratio is less than the predetermined difference ratio threshold, then supporting association label is set for the curve segment.

[0043] It is understandable that there may be multiple micro-observation time domains within a curve segment. Therefore, the curve's change characteristics are the average of the change amplitudes and the average of the change slopes corresponding to multiple micro-observation time domains. For a single micro-observation time domain, the change amplitude is the difference between the highest and lowest amplitudes, and the change slope is the average slope of the corresponding curve segment in the micro-observation time domain.

[0044] The difference ratio threshold is predetermined. In practice, several video frames belonging to the feature-dense type and with normal traffic flow and no accidents are used as samples. A time-domain change curve is constructed based on the original bitstream, and the corroborating correlation relationship of the curve segments is constructed. The mean difference ratio of curve change features between curve segments with corroborating correlation relationships is analyzed to reflect the level of difference of curve change features under normal conditions. In practice, the difference ratio threshold is set as the product of the mean difference ratio of curve change features and the adjustment coefficient. The adjustment coefficient is selected in the range [0.75, 0.95], preferably 0.85, in order to appropriately reduce the mean difference ratio, improve sensitivity, and determine that the curve segments with corroborating correlation labels have similar performance patterns.

[0045] This invention adapts to the category of curve segments through corroborative correlation analysis. For curve segments with dense features, the micro-observation time domain is determined based on inflection points. In practice, for curve segments with dense features, changes are more pronounced at the moments when vehicles enter and exit, easily forming inflection points. Since traffic flow changes are non-transient, these changes exhibit certain regularities and similarities under normal vehicle conditions. Therefore, a micro-observation time domain is constructed to observe the curve change characteristics within the time domain where changes are more pronounced. This invention does not employ a method that captures obvious abrupt changes; under multi-traffic conditions, the time-domain curve fluctuations are significant, and such methods may be affected by... To address the issue of low accuracy due to the sporadic nature of image occurrences, this invention considers curve segments with corroborating relationships that can be used for mutual verification when no anomalies are present. By comparing the curve variation characteristics of these segments, mutual verification can be achieved, ensuring high accuracy and rapid setting of corroborating relationship labels with minimal computational effort. Furthermore, for curve segments that are not feature-dense and have small or no fluctuations, corroborating relationship curve segments are used to quickly set corroborating relationship labels by comparing similarity. This facilitates subsequent adaptive decoding analysis of the curve segments with corroborating relationship labels, thereby reducing the number of decoding frame verifications and ensuring the reliability and efficiency of anomaly identification.

[0046] Specifically, the analysis module extracts each curve segment and places them in the same coordinate system. Based on the overlap of the curve segments, it sets supporting association labels for the curve segments, including... The remaining curve segments are trimmed according to the time length corresponding to the smallest curve segment so that each curve segment corresponds to the same time length. Place all the trimmed curve segments in the same coordinate system and calculate the degree of overlap between the curve segments; If the overlap is greater than or equal to a predetermined overlap threshold, a supporting association label is set; if the overlap is less than the predetermined overlap threshold, no supporting association label is set.

[0047] In practice, the degree of overlap is the ratio of the overlapping area to the average area of ​​the two curve segments.

[0048] In practice, the overlap threshold is preset. Curve segments with corroborating correlations collected by the same acquisition terminal are obtained in advance, and the overlap is calculated separately. The normal distribution of the overlap is determined, and the lower limit of the 95% confidence interval is set as the overlap threshold to reflect situations where the overlap deviates from the normal.

[0049] For details, please refer to Figure 4 As shown, it is a logic block diagram of the original bitstream in each time domain interval decoded and analyzed based on the corroborating association tag in an embodiment of the invention. The intelligent processing module extracts the original bitstream to obtain the decoding frame and determine the verification tag according to the category of the curve segment for the set corroborating association tag, and assigns the verification tag to all the original bitstream in the time domain interval. For curve segments without supporting labels, time-series decoding is performed on the original bitstream to obtain decoded frames, and the verification labels of the original bitstream are determined sequentially based on the decoded frames.

[0050] This invention decodes and analyzes the original bitstreams in each time domain interval based on corroborating association tags. For the curve segments with corroborating association tags, it reflects that there are no potential anomalies after corroborating association analysis. Therefore, subsequently, after decoding a local original bitstream, the verification tag is determined and the verification tag is assigned to all original bitstreams in the time domain interval to ensure the accuracy of identification and significantly reduce the number of decoding frames involved in the verification.

[0051] Specifically, the intelligent processing module determines the verification tag corresponding to the original bitstream based on the decoded frame, including: Determine whether the predetermined target exists in the decoded frame; If a predetermined target exists, set a verification tag for the original bitstream corresponding to the decoded frame; If no predetermined target exists, do not set a verification tag for the original bitstream corresponding to the decoded frame.

[0052] Specifically, there are no restrictions on the method of verifying the decoded frames. For example, existing target detection models can be used to identify whether there are predetermined targets in the image, such as piles of objects, or gatherings of vehicles and people. Alternatively, an image processing module that can identify traffic accident anomalies can be trained independently. This invention reduces the number of decoded frames by performing pre-analysis with extremely low computational power consumption, thereby reducing the input to the model and only inputting decoded frames with potential anomalies, thus improving analysis efficiency.

[0053] Specifically, the verification tag is a virtual tag, and there are no restrictions on how it is set. Based on the modification of the original bitstream, a digital tag can be constructed and associated with the corresponding original bitstream. Of course, other forms are also possible, which will not be elaborated here.

[0054] Specifically, it also includes a response module, which responds to the setting of a verification tag on the original bitstream, and is used to lock the acquisition terminal based on the original bitstream and perform anomaly handling, including... Based on the correlation between the data acquisition terminal and location information, abnormal locations are determined, and the abnormal locations and corresponding decoded frames are sent to the processing terminal. Simultaneously, the abnormal locations are pushed to the corresponding roadside display boards.

[0055] Specifically, the location of the data collection terminal is relatively fixed. Therefore, the association between the data collection terminal and the location information can be pre-established. Knowing which data collection terminal the original code stream of the verified tag belongs to can locate the abnormal location. The processing end is the computer or mobile terminal of the emergency response personnel for emergency response dispatch.

[0056] In some possible implementations, roadside display boards are installed on certain road sections. After an abnormal location is identified, the corresponding information can be displayed on the roadside display board, such as "An accident has occurred on a certain road section. Please pay attention to safety." This will not be elaborated further.

[0057] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. An event-response-based highway management system, characterized in that, include: The multi-source sensing module is used to acquire the raw bitstreams collected by several acquisition terminals and extract coding domain features based on the raw bitstreams to construct time-domain variation curves. The correlation processing module is used to segment the time-domain change curves based on curve features, classify the obtained curve segments, and group and associate them according to the time sequence of the same type of curve segments to build corroborative correlation relationships between each curve segment. The analysis module is used to identify curve segments with corroborating relationships, and performs corroborating relationship analysis based on the category corresponding to the curve segment, including: The inflection points in the calibrated curve segments are determined, and several micro-observation time domains are determined based on each inflection point to extract the curve change characteristics within each micro-observation time domain. Based on the differences in curve change characteristics between curve segments, corroborating association labels are set for the curve segments. Each curve segment is extracted and placed in the same coordinate system. Based on the overlap of the curve segments, corroborating association labels are set for the curve segments. The intelligent processing module is used to determine the time domain interval corresponding to each curve segment, and to decode and analyze the original bitstream within each time domain interval based on supporting correlation tags, including: Based on the category of the curve segment, the original bitstream is extracted to obtain the decoded frame and determine the verification label. The verification label is then assigned to all the original bitstreams within the time domain interval. The original bitstream is time-series decoded to obtain decoded frames, and the verification tags of the original bitstream are determined sequentially based on the decoded frames.

2. The event-response-based highway management system according to claim 1, characterized in that, The multi-source sensing module extracts coding domain features from the original code stream to construct a time-domain variation curve, including... The proportion of target macroblocks and the macroblock division depth of the corresponding video frames at each time point are determined based on the original bitstream. The coding domain features are obtained by normalizing the proportion of the target macroblock and the macroblock partitioning depth, and then summing them by weight. A time-domain variation curve is constructed with the coding domain features as the vertical axis and time as the horizontal axis. The target macroblock is a non-skipped macroblock.

3. The event-response-based highway management system according to claim 1, characterized in that, The correlation processing module is used to segment time-domain variation curves based on curve features, including... Construct time windows and slide them on the time axis to determine the curve characteristics of each time window, including the floating frequency and the average amplitude; Verify the rate of change of the curve features corresponding to the adjacent time windows in the verification time window, lock the segmented time window based on the rate of change, and determine the first moment of the segmented time window as the segmentation moment; The time-domain variation curve is divided into several curve segments based on the segmentation time. If the rate of change of a curve feature is greater than the predetermined rate of change of the segmentation threshold, then the time window is locked as the segmentation time window.

4. The event-response-based highway management system according to claim 1, characterized in that, The association processing module is used to classify the obtained curve segments, group and associate them according to the temporal order of curve segments of the same type, and construct corroborating association relationships between each curve segment, including... If a curve segment meets the floating condition, the curve segment is classified as a feature-dense type. If the curve segment does not meet the floating condition, the curve segment is classified as a non-feature dense type. Identify temporally adjacent curve segments of the same type, verify the maximum time interval between curve segments of the same type, and if the maximum time interval is less than a predetermined interval threshold, determine that there is a corroborating correlation between curve segments of the same type. The floating condition is that the floating frequency of the curve segment is greater than a predetermined floating frequency threshold.

5. The event-response-based highway management system according to claim 1, characterized in that, The analysis module performs corroborating correlation analysis based on the category corresponding to the curve segment, wherein... For curve segments with dense features, inflection points in the calibrated curve segments are selected, and several micro-observation time domains are determined based on each inflection point to extract curve change features within each micro-observation time domain. Based on the differences in curve change features between curve segments, corroborating association labels are set for the curve segments. For curve segments that are not feature-dense, select each curve segment and place it in the same coordinate system. Based on the overlap of the curve segments, set corroborating association labels for the curve segments.

6. The event-response-based highway management system according to claim 1, characterized in that, The analysis module extracts the curve change features within each micro-observation time domain, and sets corroborating association labels for the curve segments based on the differences in curve change features between them. Determine the intermediate time between each inflection point and other adjacent inflection points, and define the time interval formed by two intermediate times as the micro-observation time domain; Extract the curve change characteristics corresponding to the curve segment within the micro-observation time domain. This includes the average magnitude of the change and the average slope of the change; Compare the curve change characteristics between curve segments to determine the difference ratio of curve change characteristics; Specifically, if the curve change feature difference ratio is greater than or equal to a predetermined difference ratio threshold, no supporting association label is set for the curve segment; if the curve change feature difference ratio is less than the predetermined difference ratio threshold, then supporting association label is set for the curve segment.

7. The event-response-based highway management system according to claim 1, characterized in that, The analysis module extracts each curve segment and places them in the same coordinate system. Based on the overlap of the curve segments, it sets corroborating association labels for the curve segments, including... The remaining curve segments are trimmed according to the time length corresponding to the smallest curve segment so that each curve segment corresponds to the same time length. Place all the trimmed curve segments in the same coordinate system and calculate the degree of overlap between the curve segments; If the overlap is greater than or equal to a predetermined overlap threshold, a supporting association label is set; if the overlap is less than the predetermined overlap threshold, no supporting association label is set.

8. The event-response-based highway management system according to claim 1, characterized in that, The intelligent processing module extracts the original bitstream and obtains the decoding frame to determine the verification label for the curve segment with the set corroborating association label, according to the category of the curve segment, and assigns the verification label to all the original bitstreams in the time domain interval. For curve segments without supporting labels, time-series decoding is performed on the original bitstream to obtain decoded frames, and the verification labels of the original bitstream are determined sequentially based on the decoded frames.

9. The event-response-based highway management system according to claim 1, characterized in that, The intelligent processing module determines the verification tags corresponding to the original bitstream based on the decoded frames, including: Determine whether the predetermined target exists in the decoded frame; If a predetermined target exists, set a verification tag for the original bitstream corresponding to the decoded frame; If no predetermined target exists, do not set a verification tag for the original bitstream corresponding to the decoded frame.

10. The event-response-based highway management system according to claim 1, characterized in that, It also includes a response module, which responds to the setting of a verification tag on the original bitstream, and is used to lock the acquisition terminal based on the original bitstream and perform anomaly handling, including... Based on the correlation between the data acquisition terminal and location information, abnormal locations are determined, and the abnormal locations and corresponding decoded frames are sent to the processing terminal. Simultaneously, the abnormal locations are pushed to the corresponding roadside display boards.

Images