Multi-video stream color change intelligent monitoring method and system based on multi-scale time sequence analysis

Abstract

The invention discloses a multi-video stream color change intelligent monitoring method and system based on multi-scale time sequence analysis, and belongs to the technical field of intelligent video monitoring. According to the method, aiming at the problem that in sewage treatment and other scenes, abnormal red turning of a water body needs to be monitored in real time, firstly, an independent processing channel is established for each video stream, and a current frame is converted into an HSV color space from an RGB color space; a double-range HSV red detection algorithm is adopted to accurately identify a red area, and the red pixel proportion of each frame is calculated; performing time sequence fusion on the proportion sequence through moving average filtering, and dynamically updating a reference red proportion based on a weighted sliding window; constructing a multi-scale criterion to realize anomaly judgment by combining a self-adaptive absolute threshold and a relative change rate threshold; once a remarkable color mutation is detected, an alarm is triggered immediately, and information such as a timestamp, a red proportion and a change rate is output. The system supports multi-channel video concurrent processing, has high sensitivity, strong robustness and low false alarm rate, and is suitable for intelligent visual monitoring in the fields of water affairs, environmental protection and the like.

Description

Technical Field

[0001] This invention belongs to the field of computer vision and intelligent monitoring technology, and specifically relates to an intelligent monitoring method and system for color changes in multiple video streams based on multi-scale temporal analysis. Background Technology

[0002] With the continuous improvement of industrial automation and intelligence, higher demands are being placed on the intelligent analysis capabilities of video surveillance systems. Traditional video surveillance systems mainly rely on manual observation to detect anomalies, which suffers from low efficiency, fatigue, and high false negative rates. Especially in applications requiring long-term monitoring of specific color changes, such as liquid leak detection in chemical plants and color change monitoring of test strips in laboratories, traditional methods struggle to meet the requirements for real-time performance and accuracy.

[0003] Existing color change detection technologies mainly suffer from the following problems: Poor environmental adaptability: Environmental factors such as changes in lighting and minor adjustments to the camera position can easily lead to false alarms or missed detections; High noise sensitivity: Single-frame detection is easily affected by factors such as video noise and network jitter; Lack of time series analysis: Time series information is not fully utilized for robustness testing; Threshold setting is difficult: fixed thresholds are hard to adapt to different scenarios and environmental conditions; Limited multi-stream processing capabilities: It cannot effectively manage and coordinate the concurrent detection of multiple video streams.

[0004] Therefore, there is an urgent need to develop a color change monitoring system with strong environmental adaptability, high noise resistance, intelligent threshold adjustment, and multi-stream concurrent processing capabilities. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for intelligent monitoring of color changes in multiple video streams based on multi-scale time-series analysis, so as to overcome the defects of the prior art, such as poor environmental adaptability, high noise sensitivity, lack of time-series analysis capability and low multi-stream processing efficiency.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for intelligent monitoring of color changes in multiple video streams based on multi-scale temporal analysis includes the following steps: S1. Initialize multiple monitoring threads and establish an independent processing channel for each video stream; S2. Perform RGB to HSV color space conversion on the current frame of each video stream; S3. Red region detection is performed based on the dual-range HSV red detection algorithm, wherein the dual range covers the low-tone and high-tone red regions in the color wheel; S4. Calculate the percentage of red in each frame; S5. Perform a moving average filter on the red proportion sequence to obtain the fused red proportion; S6. Dynamically update the baseline red proportion based on a weighted sliding window; S7. Combine the adaptive absolute threshold and the adaptive relative rate of change threshold to perform multi-scale color change criteria to determine whether an anomaly has occurred. S8. When an anomaly is detected, an alarm mechanism is triggered and an anomaly message is output.

[0007] Preferably, in step S2, the mathematical expression for the RGB to HSV color space conversion is: ; in, It is the input RGB image. It is the output HSV image. Represents pixel coordinates, Represents a timestamp.

[0008] Preferably, in step S3, the dual-range HSV red detection algorithm defines two detection masks. The red detection mask is the result of a logical OR operation of the two masks. The first mask covers the region with hue H∈[0,10], saturation S∈[70,255], and lightness V∈[50,255]. The second mask covers the region with hue H∈[170,180], saturation S∈[70,255], and lightness V∈[50,255].

[0009] Preferably, in step S3, the dual-range HSV red detection algorithm defines the following two detection masks: ; ; The red detection mask is: ; H, S, and V represent the hue, saturation, and brightness channels of the HSV format, respectively.

[0010] Preferably, in step S4, the proportion of red in the current frame is obtained by counting the total number of red pixels and dividing it by the total number of pixels in the image.

[0011] Preferably, in step S4, the formula for calculating the proportion of red in the current frame is: ; in, This represents the total number of red pixels in the current frame. and These are the image's width and height, respectively.

[0012] Preferably, in step S5, the red proportion after fusion is obtained by calculating the arithmetic average of the red proportions of the most recent L frames.

[0013] Preferably, in step S5, the proportion of red after fusion is obtained by moving average filtering, and its calculation formula is as follows: ; in, This represents the percentage of red after the moving average filtering. This represents the length of the filtering window.

[0014] Preferably, in step S6, the baseline red proportion is obtained by collecting the red proportions of the previous N frames and calculating a weighted average according to the exponential decay weight.

[0015] Preferably, in step S6, the initial value of the baseline red proportion is obtained by weighted average calculation, and the formula is: ; in, For the first The weight coefficients of the frame, and satisfying: ; parameter Control the rate of time decay; the baseline value is updated under timed or continuous no-abnormal conditions, and the update formula is: ; in, To update the weights, This represents the average percentage of red within the most recent time window.

[0016] Preferably, in step S7, the relative change rate is calculated using the relative change rate when the proportion of the base red is greater than the preset minimum value, and the absolute change rate is calculated when the proportion of the base red is less than or equal to the preset minimum value; the multi-scale color change criterion is determined to be abnormal when the proportion of red after fusion exceeds the absolute threshold or the relative change rate exceeds the relative change rate threshold.

[0017] Preferably, in step S7, the relative rate of change The calculation formula is: ; in, Minimal positive numbers to prevent division by zero errors; multi-scale color change criteria. The expression is: ; in, For absolute threshold, This is the threshold for the relative rate of change.

[0018] Preferably, the adaptive absolute threshold in step S7 and adaptive relative threshold The calculation formulas are as follows, based on dynamic adjustments using historical data statistics: ; ; in, This represents the historical average percentage of red. Standard deviation and To adjust the coefficient, The median absolute deviation of the rate of change series is defined as: ; Preferably, in step S8, the abnormal information includes a timestamp, the proportion of red, the rate of change, and the alarm status.

[0019] A multi-video stream color change intelligent monitoring system based on multi-scale temporal analysis is provided. The system employs the aforementioned multi-video stream color change intelligent monitoring method based on multi-scale temporal analysis. The system includes: The multi-stream management module is used for concurrent access and connection management of multiple video streams; The color space conversion module, connected to the multi-stream management module, is used to perform RGB to HSV color space conversion; The red region detection module is connected to the color space conversion module and is used to detect red regions based on the dual-range HSV red detection algorithm. The red proportion calculation module is connected to the red region detection module and is used to calculate the red proportion of each frame; The time-series fusion processing module is connected to the red proportion calculation module and is used to perform moving average filtering on the red proportion sequence; An adaptive benchmark update module, connected to the time-series fusion processing module, is used to dynamically update the benchmark red proportion based on a weighted sliding window. The intelligent threshold optimization module is used to dynamically adjust the absolute threshold and the relative rate of change threshold based on historical data statistics. The anomaly detection and alarm module is connected to the time-series fusion processing module, the adaptive benchmark update module, and the intelligent threshold optimization module, respectively, and is used to execute multi-scale criteria and trigger alarms.

[0020] Preferably, the red area detection module adopts a dual-range HSV red detection algorithm, which covers the entire red area by combining the hue ranges [0,10] and [170,180].

[0021] Preferably, the adaptive benchmark update module uses a sliding window algorithm with exponentially decaying weights to update the benchmark value.

[0022] Preferably, the anomaly detection and alarm module combines absolute threshold and relative change rate as dual criteria for color change detection.

[0023] Preferably, the multi-stream management module adopts a multi-threaded architecture, with each video stream corresponding to an independent monitoring thread, and the detection results of each stream are stored uniformly through shared memory or message queue.

[0024] A computer device includes one or more processors, on which one or more executable programs are stored, wherein when the one or more executable programs are executed by the one or more processors, they are used to implement the intelligent monitoring method for color changes of multiple video streams based on multi-scale time-series analysis.

[0025] The present invention can achieve the following beneficial effects: 1. The adaptive baseline update mechanism of this invention can automatically adapt to environmental factors such as changes in lighting and fine adjustments to the camera position, thereby significantly reducing the false alarm rate; 2. The multi-frame fusion and temporal filtering techniques of this invention effectively suppress the effects of video noise and network jitter, improving detection stability; 3. The adaptive threshold optimization algorithm of this invention, based on historical data statistics, can automatically adjust the detection parameters according to the actual application scenario; 4. The multi-threaded architecture of this invention supports simultaneous monitoring of multiple video streams, significantly improving system processing efficiency; 5. The optimized algorithm and concurrent processing mechanism of this invention ensure that the system can respond to color change events in real time; 6. The modular design of this invention facilitates system expansion and maintenance, and supports different types of color detection tasks. Attached Figure Description

[0026] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0027] Preferred solutions include Figure 1 As shown, a method and system for intelligent monitoring of color changes in multiple video streams based on multi-scale temporal analysis specifically includes the following: S1. Initialize multiple monitoring threads and establish an independent processing channel for each video stream; S2. Perform RGB to HSV color space conversion on the current frame of each video stream; Input RGB image Convert to HSV color space:

[0028] in, This represents the conversion function from RGB to HSV color space. It is the output HSV image. Represents pixel coordinates, This indicates the timestamp. This step converts the input RGB image to HSV format, which facilitates subsequent extraction of red regions based on hue (HSV color space better matches human visual perception, and red regions are more concentrated in the H channel).

[0029] S3. Red region detection is performed based on the dual-range HSV red detection algorithm, wherein the dual range covers the low-tone and high-tone red regions in the color wheel; Based on the distribution characteristics of red in the HSV color wheel (red is mainly concentrated in...) Nearby and (Nearby), define two detection ranges:

[0030]

[0031] In this context, H, S, and V represent the three channels of the HSV format: H represents hue, S represents saturation, and V represents value. Covering low-tone red areas, Cover the high-tone red areas.

[0032] The red detection mask is:

[0033] Using an OR operation indicates that the red detection mask covers all the red areas in the HSV color wheel.

[0034] S4. Calculate the percentage of red in each frame; The formula for calculating the proportion of red in the current frame is:

[0035] in, This represents the total number of red pixels in the current frame. and These are the image's width and height, respectively.

[0036] S5. Perform a moving average filter on the red proportion sequence to obtain the fused red proportion; Robust Detection Through Multi-Frame Fusion: To improve the robustness of detection, this invention employs a multi-frame fusion strategy. Timing filters: Apply a moving average filter to the red percentage sequence:

[0037] in, This represents the percentage of red after the moving average filtering. This is the length of the filtering window. It is time The percentage of red at any given moment.

[0038] S6. Dynamically update the baseline red proportion based on a weighted sliding window; This invention proposes an adaptive benchmark update mechanism based on sliding window and weighted average: Baseline red percentage calculation: After the system starts up, the percentage of red in the first N frames of images is collected. The initial baseline is obtained by calculating a weighted average based on the weights:

[0039] in, It represents the baseline percentage of red at timestamp t, reflecting the level of red percentage under normal conditions. The percentage of red at time ti. Let be the weight coefficients for the i-th frame, satisfying:

[0040] parameter The rate of time decay is controlled, with newer frames receiving higher weights.

[0041] Benchmark update strategy: The baseline value is updated under the following conditions: Scheduled updates: every specified time interval .

[0042] Condition update: When consecutive detections are made When there are no abnormalities in the frame.

[0043] The updated baseline value is calculated as follows:

[0044] in, For the updated weighting percentages, The weighting percentages before the update. To update the weights, This represents the average percentage of red within the most recent time window.

[0045] S7. Combine the adaptive absolute threshold and the adaptive relative rate of change threshold to perform multi-scale color change criteria to determine whether an anomaly has occurred. Calculation of rate of change: The formula for calculating the rate of change relative to the benchmark is:

[0046] in, It is the rate of change of the current proportion of red relative to the benchmark. For a very small positive number to prevent division by zero error (e.g.) ).

[0047] when hour, The calculation method is a relative rate of change, which reflects the fluctuation ratio of the red proportion relative to the benchmark.

[0048] when hour, The calculation method is based on absolute rate of change to avoid calculation errors when the baseline is too small.

[0049] Multi-scale detection criteria: The comprehensive criteria for color change detection are:

[0050] in, This is an abnormal detection result. For absolute threshold, This is the threshold for the relative rate of change.

[0051] A value of 1 indicates an abnormality, while a value of 0 indicates normality. This indicates that the red area represents an excessively high percentage. This indicates that the red area has changed too much.

[0052] This invention proposes an adaptive threshold optimization method based on historical data statistics: Statistical feature extraction: Regularly collect historical red percentage sequences and rate of change series Calculate the statistical characteristics of the historical red proportion sequence: Mean: (Calculate the expected value of the historical red percentage to reflect the average level of red percentage in normal scenarios).

[0053] Standard deviation: (Standard deviation of the historical red percentage reflects the degree of fluctuation in the red percentage).

[0054] 95th percentile: (95th percentile of the historical percentage of red).

[0055] Adaptive threshold calculation: The adaptive update formula for the absolute threshold is:

[0056] in, To adapt to the absolute threshold, it is dynamically adjusted based on historical data. This is the absolute threshold adjustment factor (controlling the distance between the threshold and the mean). The larger the threshold, the more lenient it is.

[0057] The adaptive update formula for the relative threshold is as follows:

[0058] in, To adapt to the relative threshold, it is dynamically adjusted according to the historical rate of change. This is the relative threshold adjustment coefficient. The median absolute deviation of the rate of change series reflects Robustness to dispersion (resistance to extreme value disturbances):

[0059] Application of adaptive threshold: Adaptive threshold and Replace fixed thresholds for multi-scale detection criteria The calculation improves the detection adaptability in different scenarios.

[0060] This invention employs a multi-threaded concurrent processing architecture, with each video stream corresponding to an independent monitoring thread: When the system starts, it creates N independent monitoring threads, each thread corresponding to one video stream. Threads are isolated from each other but share system configurations.

[0061] Flow management strategy: Assume the system processes simultaneously The video stream, the first The processing flow for each stream is as follows: Stream connection management: Periodically check the stream connection status and automatically reconnect when disconnected. Each stream periodically (e.g., every 3 seconds) sends a heartbeat check. If there is no response for 3 consecutive times, it is determined to be disconnected and the reconnection mechanism is automatically triggered (maximum 5 retries, with an interval of 10 seconds).

[0062] Resource allocation: Computing resources are dynamically allocated based on stream priority and system load. Resources are dynamically allocated based on stream priority (high priority, such as key security areas) and system load (CPU / memory utilization). High-priority streams are allocated computing cores first, and the processing frame rate of low-priority streams is reduced when the load exceeds 80%.

[0063] Synchronization and coordination: Ensure that the processing results of each stream can be managed and displayed uniformly. This is achieved by uniformly storing the detection results (timestamps, etc.) of each stream through shared memory or message queues. (Alarm status), which is summarized and displayed by the main process.

[0064] Performance optimization model: The overall system performance index is defined as follows:

[0065] in, For the overall system performance indicators, For the first The processing performance of each stream These are the weighting coefficients.

[0066] Formula for calculating the single-stream processing performance of the i-th stream:

[0067] in: Number of frames processed in the i-th path : Total number of input frames in the i-th path (reflecting processing completeness).

[0068] : False alarm rate of the i-th channel (number of false alarms / total number of alarms).

[0069] : The false negative rate of the i-th path (number of undetected anomalies / actual number of anomalies).

[0070] S8. When an anomaly is detected, an alarm mechanism is triggered and an anomaly message is output.

[0071] The implementation flow of this invention system, i.e., the algorithm flow: Input: Multiple video streams {S1,S2,...,S} n (Supports protocols such as RTSP / HTTP).

[0072] Output: Real-time color change detection results for each stream ( (Value, red percentage curve) and abnormal alarm information (time stamp, abnormal type).

[0073] 1. System initialization (main process): Create N monitoring threads (one-to-one with the video stream) and configure thread priorities.

[0074] Initialize the baseline parameters: N (number of baseline frames). (Time decay), L (Filter window) (Timed update interval), k (number of consecutive frames without abnormalities), etc.

[0075] Initialize threshold parameters: , And pre-calculate the initial value based on historical data. and .

[0076] 2. For each video stream S i Concurrent execution: a. Connecting to the video stream: After the thread starts, it attempts to connect to the corresponding video stream S. i It automatically reconnects and logs when the connection is lost.

[0077] b. Establish initial baseline values: After successful connection, calculate the initial baseline by acquiring the first N frames.

[0078] c. Enter the main monitoring loop: While the system runs do: Read the current frame according to the frame rate (e.g., 25fps, read one frame every 40ms).

[0079] Perform RGB to HSV color space conversion to generate .

[0080] Calculate the red mask Count the number of red pixels and calculate .

[0081] Moving average fusion: Take the most recent L frames calculate .

[0082] Calculate the rate of change Combined with adaptive threshold determination (Abnormal / Normal).

[0083] Benchmark update check: If the timed update is met Or if there are no anomalies for k consecutive frames, update. .

[0084] Result storage: Store t, R(t), , , Write to the database.

[0085] Alarm triggered: If .

[0086] Endwhile.

[0087] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for intelligent monitoring of color changes in multiple video streams based on multi-scale temporal analysis, characterized in that, Includes the following steps: S1. Initialize multiple monitoring threads and establish an independent processing channel for each video stream; S2. Perform RGB to HSV color space conversion on the current frame of each video stream; S3. Red region detection is performed based on the dual-range HSV red detection algorithm, wherein the dual range covers the low-tone and high-tone red regions in the color wheel; S4. Calculate the percentage of red in each frame; S5. Perform a moving average filter on the red proportion sequence to obtain the fused red proportion; S6. Dynamically update the baseline red proportion based on a weighted sliding window; S7. Combine the adaptive absolute threshold and the adaptive relative rate of change threshold to perform multi-scale color change criteria to determine whether an anomaly has occurred. S8. When an anomaly is detected, an alarm mechanism is triggered and an anomaly message is output.

2. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 1, characterized in that, In step S2, the mathematical expression for the RGB to HSV color space conversion is: ； in, It is the input RGB image. It is the output HSV image. Represents pixel coordinates, Represents a timestamp.

3. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 1, characterized in that, In step S3, the dual-range HSV red detection algorithm defines two detection masks. The red detection mask is the result of a logical OR operation of the two masks. The first mask covers the region with hue H∈[0,10], saturation S∈[70,255], and lightness V∈[50,255]. The second mask covers the region with hue H∈[170,180], saturation S∈[70,255], and lightness V∈[50,255].

4. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 3, characterized in that, In step S3, the dual-range HSV red detection algorithm defines the following two detection masks: ； ； The red detection mask is: ； H, S, and V represent the hue, saturation, and brightness channels of the HSV format, respectively.

5. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 1, characterized in that, In step S4, the proportion of red in the current frame is obtained by counting the total number of red pixels and dividing it by the total number of pixels in the image.

6. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 1, characterized in that, In step S4, the formula for calculating the proportion of red in the current frame is: ； in, This represents the total number of red pixels in the current frame. and These represent the width and height of the image, respectively.

7. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 1, characterized in that, In step S5, the red proportion after fusion is obtained by calculating the arithmetic average of the red proportions of the most recent L frames.

8. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 7, characterized in that, In step S5, the proportion of red after fusion is obtained through moving average filtering, and its calculation formula is as follows: ； in, This represents the percentage of red after the moving average filtering. This represents the length of the filtering window.

9. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 1, characterized in that, In step S6, the baseline red proportion is obtained by collecting the red proportions of the previous N frames and calculating a weighted average based on the exponential decay weight.

10. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 1, characterized in that, In step S6, the initial value of the baseline red proportion is calculated by weighted average, and the formula is as follows: ； in, For the first The weight coefficients of the frame, and satisfying: ； parameter Control the rate of time decay; the baseline value is updated under timed or continuous no-abnormal conditions, and the update formula is: ； in, To update the weights, This represents the average percentage of red within the most recent time window.

11. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 1, characterized in that, In step S7, the relative change rate is calculated when the proportion of the base red is greater than the preset minimum value, and the absolute change rate is calculated when the proportion of the base red is less than or equal to the preset minimum value; the multi-scale color change criterion is determined to be abnormal when the proportion of red after fusion exceeds the absolute threshold or the relative change rate exceeds the relative change rate threshold.

12. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 11, characterized in that, In step S7, the relative rate of change The calculation formula is: ； in, Minimal positive numbers to prevent division by zero errors; multi-scale color change criteria. The expression is: ； in, For absolute threshold, This is the threshold for the relative rate of change.

13. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 12, characterized in that, The adaptive absolute threshold in step S7 and adaptive relative threshold The calculation formulas are as follows, based on dynamic adjustments using historical data statistics: ； ； in, This represents the historical average percentage of red. Standard deviation and To adjust the coefficient, The median absolute deviation of the rate of change series is defined as: 。 14. The intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to claim 1, characterized in that, In step S8, the abnormal information includes timestamp, red percentage, change rate, and alarm status.

15. A multi-video stream color change intelligent monitoring system based on multi-scale temporal analysis, characterized in that, The system employs the intelligent monitoring method for color changes in multiple video streams based on multi-scale temporal analysis according to any one of claims 1-14, and includes: The multi-stream management module is used for concurrent access and connection management of multiple video streams; The color space conversion module, connected to the multi-stream management module, is used to perform RGB to HSV color space conversion; The red region detection module is connected to the color space conversion module and is used to detect red regions based on the dual-range HSV red detection algorithm. The red proportion calculation module is connected to the red region detection module and is used to calculate the red proportion of each frame; The time-series fusion processing module is connected to the red proportion calculation module and is used to perform moving average filtering on the red proportion sequence; An adaptive benchmark update module, connected to the time-series fusion processing module, is used to dynamically update the benchmark red proportion based on a weighted sliding window. The intelligent threshold optimization module is used to dynamically adjust the absolute threshold and the relative rate of change threshold based on historical data statistics. The anomaly detection and alarm module is connected to the time-series fusion processing module, the adaptive benchmark update module, and the intelligent threshold optimization module, respectively, and is used to execute multi-scale criteria and trigger alarms.

16. The intelligent monitoring system for color changes in multiple video streams based on multi-scale temporal analysis according to claim 15, characterized in that, The red area detection module uses a dual-range HSV red detection algorithm, which combines hue ranges [0,10] and [170,180] to cover the entire red area.

17. The intelligent monitoring system for color changes in multiple video streams based on multi-scale temporal analysis according to claim 15, characterized in that, The adaptive benchmark update module uses a sliding window algorithm with exponentially decaying weights to update the benchmark value.

18. The intelligent monitoring system for color changes in multiple video streams based on multi-scale temporal analysis according to claim 15, characterized in that, The anomaly detection and alarm module combines absolute threshold and relative change rate as dual criteria to detect color changes.

19. The intelligent monitoring system for color changes in multiple video streams based on multi-scale temporal analysis according to claim 15, characterized in that, The multi-stream management module adopts a multi-threaded architecture, with each video stream corresponding to an independent monitoring thread, and the detection results of each stream are stored uniformly through shared memory or message queue.

20. A computer device, characterized in that, The method includes one or more processors, on which one or more executable programs are stored, and when the one or more executable programs are executed by the one or more processors, they are used to implement the intelligent monitoring method for color changes of multiple video streams based on multi-scale time-series analysis according to any one of claims 1-14.

Images