Product

Position : HillSunTec > Products > Image quality detection

Image quality detection

 

Video quality diagnosis is an intelligent video fault analysis and warning system. For video image of snowflakes, scrolling, fuzzy, partial color, screen freezes, gain unbalance, video signal loss, monitoring camera image shift , video signal interference, the decrease of the quality of the video for accurate analysis, judgment, and report to the police. Video quality diagnostic system can also be used as a major alarm means for preventing the degradation of video quality caused by malicious damage to the front-end camera and the loss of video.

图像质量检测.jpg

System features

Video color interference detection:Based on the core algorithm of image detection, it is possible to detect the image bias interference in the real time of front camera.

It includes follows steps:

Convert the image of color space from RGB to Lab.

Calculate the color histogram of a and b components.

Calculate the image partial color factor. The bias factor is larger than the set threshold image, which is considered to be biased and alarming.

Video snow interference detection:

Snowflake screen is a large number of irregular movement noise, and flicker in  the picture. These characteristics are: The texture is deep and regular. The grayscale value of “flicker point” is changed  by dramatically and the amplitude fluctuates greatly.

Based on the core algorithm of image detection, it can detect whether the front camera has the phenomenon of snowflake interference in real time. If there is an exception, the alarm will be alerted.

Video resolution anomaly detection:

Based on the core algorithm of image detection, it can detect the phenomenon of the image sharpness in the real time of front camera, and alarm prompt if there is any abnormality.

It includes follows steps:

Compare the resolution of quantified values and externally set up of clear warning threshold and definition alarm threshold, determines whether or not to warning or alarm the sharpness of the current image.

Compare the resolution of quantified values and externally set up of clear warning threshold and definition alarm threshold, determines whether or not to warning or alarm the sharpness of the current image include: If definition quantized value less than definition warning threshold value, then no warning or alarm information is required; If definition quantized value less than definition alarm threshold, and more than definition warning threshold value, then output warning information; If definition quantized value greater than definition alarm threshold, then output alarm information.

Video fringe interference detection:

Based on the core algorithm of image detection, it can detect the phenomenon of the fringe interference in the real time of front camera, and alarm prompt if there is any abnormality.

It includes follows steps:

First, obtain the target of video, extract the image in the video sequence, and convert the storage format of the image from multi-channel to single-channel gray level image storage format.

Secondly, the grayscale image sequence is processed the frames before and after  in order to obtain the difference frame of image sequence. And the image sequence of the difference image is de-noising processed in morphology.

Then de-noised images sequence and difference image sequence to Sobel operator sharpening processing, the two sets of images were obtained and the row projection was performed respectively. The variance of the two sets of projection data was calculated,through the variance ratio of projection data ,determine whether there are across the grain.

Picture flutter detection:

Based on image detection core algorithm, it can detect the image jitter of the front camera in real time, and alert the alarm if there is an abnormality.

It includes follows steps:

According to the current image of video YUV image information, obtain the basic feature information of the edge images of the YUV components of the current image.

Calculate the local extreme point distribution state of the edge image of the current image, compare the background image and the local extreme point distribution state of the current image, determines whether the current image is moving relative to the background image.

Video picture lose detection:

Based on image detection core algorithm, it can detect the image jitter of the front camera in real time, and alert the alarm if there is an abnormality.

It includes follows steps:

According to the video YUV image information of the current image, obtain the basic feature information of the edge images of the YUV components of the current image.

Calculate the local extreme point distribution state of the edge image of the current image, compare the background image and the local extreme point distribution state of the current image, determines whether the current image is moving relative to the background image.

Calculate the edge width of each local extreme point of the edge image of the current image, obtain the edge width average value, according the average value of the edge width to obtain the definition quantized value. determines whether the current image is moving relative to the background image, output the alarm message.

Video screen loss detection:

Based on image detection core algorithm, it can detect the image jitter of the front camera in real time, and alert the alarm if there is an abnormality.

It includes follows steps:

First, extract one frame image from the video, the analysis the gray value of each pixel of this frame image, between pixels in the same frame of difference gray value represent the intensity of the change in the frame.