There are many Background Modeling Methods! Each method has its own advantages and disadvantages! Here we will briefly introduce the background Modeling Method Based on GMM + HSV.
I. Background Modeling Algorithm
As a background modeling algorithm, when applied to real-time video monitoring, the following conditions must be met:
1) It can adapt to light changes, the leaf swing, and anti-interference.
2) It can cope with significant changes in the mean and variance of background information. To address this problem, we often use multiple models, such as Gaussian mixture Background Modeling.
3) it can handle the shadows of moving objects and detect moving objects in the shade.
4) fast computing, real-time monitoring. Currently, the common video stream is 25 FPS. Therefore, the actual processing time of an image is only 40 ms. Therefore, the simpler the processing algorithm, the better.
Ii. Background Modeling Algorithm for the background
In comparison, the Background Modeling Algorithm Based on HSV has the following advantages:
1
In this way, the image does not have to change the image's HSV value. Therefore, it can well suppress the changes in illumination and the small jitter of the camera. However, when the target is large, it will also make an error, missed detection.
2) fast computing speed. For 352*288 images, the 15mspf speed can be reached without downsampling rate. This is better than the Gaussian mixture model.
3) It can track fast moving targets in real time and has good effects on small targets.
4) It can effectively suppress noise and avoid the spot noise during gaussian background modeling. Therefore, no filtering operation is required, saving time.
5) It is complex for lighting changes, swinging trees, fluctuating lakes, and flickering monitors, there are also situations in which new objects enter or old objects are removed from the background,
6) The disadvantage is that motion shadows are not solved.
Iii. Calculation of HSV
1. computing process
The most basic HSV operator is to compare the center pixel with each pixel in its 3 × 3 neighborhood. If the gray value of the neighboring pixel is greater than or equal to the gray value of the center pixel, set the value of the neighboring pixel in the sequence of binary values of the HSV to 1; otherwise, set it to 0. Finally, an eight-bit binary number formed by sorting in a certain order is obtained by the 8 pixels in the 3 × 3 neighborhood. The decimal number of this value is used as the BPS value of this pixel, its size range is 0 ~ 255.
Recently, I am going to sort out the background modeling algorithms of the source code library and the GMM, and share them with you!