Hand posture detection with OpenCV #8
Description
简单的手势识别，基本思路是基于皮肤检测，皮肤的颜色在HSV颜色空间下与周围环境的区分度更高，从RGB转换到HSV颜色空间下针对皮肤颜色进行二值化，得到mask：
def HSVBin(img):
hsv = cv2.cvtColor(img, cv2.COLOR_RGB2HSV)
lower_skin = np.array([100, 50, 0])
upper_skin = np.array([125, 255, 255])
mask = cv2.inRange(hsv, lower_skin, upper_skin)
# res = cv2.bitwise_and(img, img, mask=mask)
return mask然后通过腐蚀与膨胀等形态学变化去除一些噪点，得到更完整的白色（皮肤）色块，最后找出色块的轮廓，并通过色块大小排除一些面积较小的噪点：
def getContours(img):
kernel = np.ones((5,5),np.uint8)
closed = cv2.morphologyEx(img, cv2.MORPH_OPEN, kernel)
closed = cv2.morphologyEx(closed, cv2.MORPH_CLOSE, kernel)
contours, h = cv2.findContours(closed, cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
validContours = [];
for cont in contours:
if cv2.contourArea(cont) > 9000:
# x,y,w,h = cv2.boundingRect(cont)
# if h/w > 0.75: # filter face failed
validContours.append(cv2.convexHull(cont))
# rect = cv2.minAreaRect(cont)
# box = cv2.cv.BoxPoints(rect)
# validContours.append(np.int0(box))
return validContours如果需要排除脸部皮肤干扰，可以同时加入人脸识别，在转换颜色空间之前将识别到的脸部位置颜色去除掉，最终运行结果如下：
🍺
Metadata
Metadata
Assignees
Projects
Milestone
Relationships
Development
No branches or pull requests
Participants
Activity