博客

附上我自己的实例代码

基于运动信息和肤色检测的手位置检测

基于深度学习的asl手语识别例程

1 椭圆肤色检测模型

原理：

将RGB图像转换到YCRCB空间，肤色像素点会聚集到一个椭圆区域。先定义一个椭圆模型，然后将每个RGB像素点转换到YCRCB空间比对是否再椭圆区域，是的话判断为皮肤。

YCRCB颜色空间

椭圆模型

代码

def ellipse_detect(image):
    """
    :param image: 图片路径
    :return: None
    """
    img = cv2.imread(image,cv2.IMREAD_COLOR)
    skinCrCbHist = np.zeros((256,256), dtype= np.uint8 )
    cv2.ellipse(skinCrCbHist ,(113,155),(23,15),43,0, 360, (255,255,255),-1)

    YCRCB = cv2.cvtColor(img,cv2.COLOR_BGR2YCR_CB)
    (y,cr,cb)= cv2.split(YCRCB)
    skin = np.zeros(cr.shape, dtype=np.uint8)
    (x,y)= cr.shape
    for i in range(0,x):
        for j in range(0,y):
            CR= YCRCB[i,j,1]
            CB= YCRCB[i,j,2]
            if skinCrCbHist [CR,CB]>0:
                skin[i,j]= 255
    cv2.namedWindow(image, cv2.WINDOW_NORMAL)
    cv2.imshow(image, img)
    dst = cv2.bitwise_and(img,img,mask= skin)
    cv2.namedWindow("cutout", cv2.WINDOW_NORMAL)
    cv2.imshow("cutout",dst)
    cv2.waitKey()

效果

2 YCrCb颜色空间的Cr分量+Otsu法阈值分割算法

原理

针对YCRCB中CR分量的处理，将RGB转换为YCRCB，对CR通道单独进行otsu处理，otsu方法opencv里用threshold

代码

def cr_otsu(image):
    """YCrCb颜色空间的Cr分量+Otsu阈值分割
    :param image: 图片路径
    :return: None
    """
    img = cv2.imread(image, cv2.IMREAD_COLOR)
    ycrcb = cv2.cvtColor(img, cv2.COLOR_BGR2YCR_CB)

    (y, cr, cb) = cv2.split(ycrcb)
    cr1 = cv2.GaussianBlur(cr, (5, 5), 0)
    _, skin = cv2.threshold(cr1,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)

    cv2.namedWindow("image raw", cv2.WINDOW_NORMAL)
    cv2.imshow("image raw", img)
    cv2.namedWindow("image CR", cv2.WINDOW_NORMAL)
    cv2.imshow("image CR", cr1)
    cv2.namedWindow("Skin Cr+OTSU", cv2.WINDOW_NORMAL)
    cv2.imshow("Skin Cr+OTSU", skin)

    dst = cv2.bitwise_and(img, img, mask=skin)
    cv2.namedWindow("seperate", cv2.WINDOW_NORMAL)
    cv2.imshow("seperate", dst)
    cv2.waitKey()

效果

3 基于YCrCb颜色空间Cr, Cb范围筛选法

原理

类似于第二种方法，只不过是对CR和CB两个通道综合考虑

代码

def crcb_range_sceening(image):
    """
    :param image: 图片路径
    :return: None
    """
    img = cv2.imread(image,cv2.IMREAD_COLOR)
    ycrcb=cv2.cvtColor(img,cv2.COLOR_BGR2YCR_CB)
    (y,cr,cb)= cv2.split(ycrcb)

    skin = np.zeros(cr.shape,dtype= np.uint8)
    (x,y)= cr.shape
    for i in range(0,x):
        for j in range(0,y):
            if (cr[i][j]>140)and(cr[i][j])<175 and (cr[i][j]>100) and (cb[i][j])<120:
                skin[i][j]= 255
            else:
                skin[i][j] = 0
    cv2.namedWindow(image,cv2.WINDOW_NORMAL)
    cv2.imshow(image,img)
    cv2.namedWindow(image+"skin2 cr+cb",cv2.WINDOW_NORMAL)
    cv2.imshow(image+"skin2 cr+cb",skin)

    dst = cv2.bitwise_and(img,img,mask=skin)
    cv2.namedWindow("cutout",cv2.WINDOW_NORMAL)
    cv2.imshow("cutout",dst)

    cv2.waitKey()

效果

4 HSV颜色空间H,S,V范围筛选法

原理

还是转换空间然后每个通道设置一个阈值综合考虑，进行二值化操作。

代码

def hsv_detect(image):
    """
    :param image: 图片路径
    :return: None
    """
    img = cv2.imread(image,cv2.IMREAD_COLOR)
    hsv=cv2.cvtColor(img,cv2.COLOR_BGR2HSV)
    (_h,_s,_v)= cv2.split(hsv)
    skin= np.zeros(_h.shape,dtype=np.uint8)
    (x,y)= _h.shape

    for i in range(0,x):
        for j in range(0,y):
            if(_h[i][j]>7) and (_h[i][j]<20) and (_s[i][j]>28) and (_s[i][j]<255) and (_v[i][j]>50 ) and (_v[i][j]<255):
                skin[i][j] = 255
            else:
                skin[i][j] = 0
    cv2.namedWindow(image, cv2.WINDOW_NORMAL)
    cv2.imshow(image, img)
    cv2.namedWindow(image + "hsv", cv2.WINDOW_NORMAL)
    cv2.imshow(image + "hsv", skin)
    dst = cv2.bitwise_and(img, img, mask=skin)
    cv2.namedWindow("cutout", cv2.WINDOW_NORMAL)
    cv2.imshow("cutout", dst)
    cv2.waitKey()

效果

示例

import cv2
import numpy as np


def ellipse_detect(image):
    """
    :param image: img path
    :return: None
    """
    img = cv2.imread(image, cv2.IMREAD_COLOR)
    skinCrCbHist = np.zeros((256, 256), dtype=np.uint8)
    cv2.ellipse(skinCrCbHist, (113, 155), (23, 15), 43, 0, 360, (255, 255, 255), -1)

    YCRCB = cv2.cvtColor(img, cv2.COLOR_BGR2YCR_CB)
    (y, cr, cb) = cv2.split(YCRCB)
    skin = np.zeros(cr.shape, dtype=np.uint8)
    (x, y) = cr.shape
    for i in range(0, x):
        for j in range(0, y):
            CR = YCRCB[i, j, 1]
            CB = YCRCB[i, j, 2]
            if skinCrCbHist[CR, CB] > 0:
                skin[i, j] = 255
    cv2.namedWindow(image, cv2.WINDOW_NORMAL)
    cv2.imshow(image, img)
    dst = cv2.bitwise_and(img, img, mask=skin)
    cv2.namedWindow("cutout", cv2.WINDOW_NORMAL)
    cv2.imshow("cutout", dst)
    cv2.waitKey()



if __name__ == '__main__':
    ellipse_detect('./test.png')