Knowledge Accumulation: Analysis of Video Quality Diagnostic Techniques

Based on a specific video image analysis algorithm, the video diagnostic technology can intelligently analyze the video images in the surveillance system. According to the logic of the algorithm, the video quality is divided into normal, signal loss, network abnormality, occlusion, freezing, video blur, video over , video too dark, color cast, signal interference, jitter and so on. Users can view front-end cameras' real-time status, receive alarms, handle alarms, and query historical information through B/S or C/S methods, and can perform various statistics based on different attributes of the camera's area, brand, type of fault, and severity of failure. analysis.

1. Abnormal brightness detection

This detection is aimed at images such as white, dark, hard to distinguish objects, screen flickering, etc. The cause of the malfunction can be generally classified as camera exposure controller failure, gain controller failure, and the camera is subject to artificial light and other issues. The detection algorithm is relatively simple. The histogram of the brightness component of the image is analyzed and the brightness distribution is analyzed accordingly. Then the empirical threshold is set to determine whether the brightness of the current image is abnormal.

2. Image blur detection

The detection of the object in the image is not clear, the image is empty, the image contrast is low, the failure causes can be summarized as the focus is not allowed, the lens covered with dust and water vapor, was artificially smeared and blocked and other issues. For its algorithm implementation, spectral analysis can be performed on the image data in the frequency domain or the image can be divided into N regions of the same size and the average contrast thereof can be counted.

3 screen freeze detection

This detection is aimed at the phenomenon that the picture is still, and the cause of the fault can usually be summarized as the problem that the camera collection end does not refresh, the line transmission fails, and the man-made stickers. For the implementation of the algorithm, one implementation method is to use the frame difference algorithm to obtain the difference between the previous and the next frames of the image. Another implementation method can extract one frame of the image in the video frame at regular intervals, and analyze the histogram, and finally statistical analysis. The histogram similarity between frames is used to derive the algorithm detection result.

4 . Loss of signal detection

This detection is aimed at blank screens, black screens, inability to obtain bitstreams, and lack of video signals. The cause of the faults can be summarized as problems such as the inability of the network to connect, and the poor or damaged video transmission lines. The implementation of the algorithm: an implementation method can do histogram statistics on the luminance channel of the image, and obtain the algorithm processing result through the analysis of the histogram result and the discreteness detection; another implementation method can binarize the image. , And then look for the largest connected area for black or white screen, and finally get the algorithm processing result by the area of â€‹â€‹the largest connected area.

5 . Color anomaly detection

This detection is based on the appearance of a single full-screen color cast on the screen, a flickering color band on the screen, and a wide range of variegated colors in the image. The cause of the malfunction can be generally attributed to the damage of the color channel of the photosensitive device and the failure of the camera's color balance algorithm. Color signals are disturbed during transmission. For this type of problem, the general processing method is to first convert the image data to a specific color space, and then analyze the specific color channels. The following uses the YUV and HIS color space as an example. When converting to the YUV color space, the UV color difference component of the image data can be statistically calculated and averaged, and a reasonable threshold value can be set to output the detection result. When converting to the HIS color space, histogram statistics can be performed on the hue component H, and then the algorithm detection result is obtained.

6. Noise interference detection

This detection is aimed at the appearance of mixed fine lines, diagonal lines, and distortions caused by screen distortion, blurring, etc., which are commonly seen as â€œsnowflakeâ€ noise. The detection of video noise is difficult and complex, mainly due to the irregularity of different types of noise, and it is difficult for a single strategy detection algorithm to completely cover different types of noise. The following uses the "snowflake" noise detection algorithm as an example. Considering that the "snow" noise is mostly salt and pepper noise, the image may be filtered first, and then the image before and after filtering is compared to obtain the detection result of the algorithm.

7. Image shake detection

This detection is based on the phenomenon of continuous shaking on the screen. The cause of the malfunction is usually the camera pole or the pan-tilt head is unstable. The detection of such problems usually requires statistical analysis of the motion information between frames of a certain length of video sequence. Based on this idea, an implementation method is to perform feature point detection on an image, and then in a certain video sequence, the running vector of the feature point is tracked, and then the algorithm detection result is obtained. In another implementation method, the image can be divided into N regions of the same size, and then the movement direction of each region in a video sequence of a certain duration is recorded, and then the detection result of the algorithm is obtained.

8. Stripe interference detection

This detection is aimed at the appearance of horizontal stripe, vertical stripe, mesh stripe and other phenomena in the screen. The cause of the fault can usually be attributed to improper grounding of the equipment, various frequency interference of the signal transmission line, improper synchronization of the sending and receiving equipment, and other issues. The algorithm for this problem is usually developed for image features of interference fringes. The following example uses the detection of horizontal fringes. The gradient fluctuations near the fringe position in the screen are large. Therefore, the gradient data of the x and y directions of the image data can be separately made to highlight the fringes. Features, then make a straight line detection on the gradient image, and then give the algorithm detection results based on the length and motion information of the detected line.

9. Black and white image detection

This detection is aimed at displaying black-and-white images (no color information) on the screen. The cause of the malfunction can be generally classified as the problem that the photosensitive device has a damaged color channel, the color balance algorithm of the camera fails, and the color signal is interfered during transmission. The algorithm is simple to implement. After converting the image data into YUV color space, the resolution of the UV component is analyzed, and compared with the set threshold of experience, the algorithm detection result can be obtained.

10. Contrast anomaly detection

This detection is aimed at the phenomenon of blurred objects and other phenomena on the screen. It is generally caused by the camera's virtual focus or lens contamination. The detection algorithm can be designed in strict accordance with the definition of image contrast. Contrast refers to the measurement of different brightness levels between the brightest white and the darkest black in a light and dark region of an image, that is, the magnitude of an image's grayscale contrast. . Therefore, the brightest white and darkest black pixel values â€‹â€‹in the light and dark regions can be counted, and the algorithm detection results are given after the difference.

11. PTZ out of control detection

This detection is based on the phenomenon that the pan/tilt cannot rotate and the pan/tilt error responds to control commands. It is generally caused by the pan/tilt mechanical failure and improper control command configuration. The design of this detection algorithm needs to be coordinated with the PTZ control command. First, the PTZ control command is sent from the diagnostic server. Then the algorithm analyzes the motion trace of the image after sending the command, and finally compares the control command with the motion trace of the image during the period of time. , which gives the algorithm detection results.

The video image quality diagnosis technology has the following functions:

1 for video surveillance system "protect escort"

The application of video diagnosis technology combined with video surveillance system, real-time monitoring of the front-end camera, effectively prevent front-end camera failure, video transmission, human damage and other factors on the video surveillance system losses, to ensure the normal and stable operation of the system. At the same time, video diagnosis technology as a video surveillance system function to improve and supplement, only need to obtain video stream information from the original system, will not affect the normal operation of existing systems.

2 Save a lot of manpower and material resources

The application of video diagnostic technology mostly covers each front-end camera in the video surveillance system one by one through the circumstance of visual inspection, and once an abnormality is found, an alarm will be issued immediately. The management personnel are freed from the arduous work, effectively relieve the stress caused by manual rapid inspections, and greatly reduce the investment in labor costs and time costs.

3 Classified statistical analysis, auxiliary operation and maintenance management

The application of video diagnosis technology combines log data management functions. Once the system detects a device failure, the system automatically generates a system status log, a failure log, etc., and provides data and picture support. Managers can export or print out the operation status report, and analyze and classify the camera brand and installation location to determine the factors that are prone to failure and carry out targeted maintenance and management, thereby greatly improving the operating efficiency of the video surveillance system.

Plush Teddy Bear

Plush Teddy Bear,Plush Brown Teddy Bear Toys,Plush Teddy Bear Toys,Plush Four-Color Teddy Bear Doll

Shenzhen Yida Toys Co. , Ltd. , https://www.ydatoy.com