Learning opencv -- hand tracking gesture tracking

In the past few days, the weather on the island has suddenly changed. After all, my unlucky child has escaped from the cold and shot birds ~~~

In the past few days, I only wrote a simple code for gesture tracking.

Principle: background difference + skin color detection.

Background difference: Take the average value of the first 30 frames, calculate the sum of the difference between the first 30 frames, and then calculate the mean value. It is detected out of the threshold value of the background average value.

Skin Color Detection: ycrcb space is used.

Two result phases and operations.

Advantages of this method: 1. It effectively solves the situation where the human face is always detected in the skin color detection results;

2. The background difference detection result is messy;

Disadvantage: the background requirement is relatively stable. The larger the contrast, the better, and the worse the robustness.

Note: Due to the difference method involving accumulative images, ensure normalization During encoding !!! Normalize

 

# Include "stdafx. H "# include <cv. h> # include