尝试实现手势控制计算机（持续跟进）

基于语言python，主要运用的包有cv2等

 

1.在python中实现摄像头的图像获取

安装opencv：

1. 安装依赖

pip install --upgrade setuptools

pip install numpy Matplotlib

2.安装opencv python版

pip install opencv-python

 

实现摄像头图像获取代码

#-*- coding:utf-8 -*-
import cv2
import time

cap = cv2.VideoCapture(0)#调取摄像头，0表示系统默认摄像头

while True:
    ret,photo = cap.read()#读取图像
    cv2.imshow('please take our photo!',photo)#将图像传送自窗口

    key = cv2.waitKey(2)#设置等待时间，若为0图像定格

    if key == ord(" "):#输入空格截图
        filename = time.strftime('%Y%m%d-%H%M%S')+".jpg"#文件名为当前时间
        cv2.imwrite(filename,photo)#保存位置

    if key == ord('q'):#输入q退出
        break

 

3.第一次尝试

想用去背景后获取二值图像，然后检测凸包的方式来识别手势，效果很差

实时去掉背景代码

#-*- coding:utf-8 -*-
import cv2
import time

cap = cv2.VideoCapture(0)#调取摄像头，0表示系统默认摄像头
#构建3*3的椭圆核
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(3,3))
#创建BackgroundSubtractorMOG2（OpenCV自带的一种背景消减方法
fgbg = cv2.createBackgroundSubtractorMOG2(detectShadows=False)

while True:
    ret,frame = cap.read()#读取图像
    fgmask = fgbg.apply(frame)
    fgmask = cv2.morphologyEx(fgmask,cv2.MORPH_OPEN,kernel)
    im,contours,hierarchy = cv2.findContours(fgmask,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)

    frame = cv2.rectangle(frame,(150,100),(300,250),(0,255,0),2)
    # for c in contours:
    #     # 计算各轮廓的周长
    #     perimeter = cv2.arcLength(c, True)
    #     if perimeter > 188:
    #         # 找到一个直矩形（不会旋转）
    #         x, y, w, h = cv2.boundingRect(c)
    #         # 画出这个矩形
    #         cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 2)

    cv2.imshow('frame', frame)
    cv2.imshow('fgmask', fgmask)

    key = cv2.waitKey(30)#设置等待时间，若为0图像定格

    if key == ord(" "):  # 输入空格截图
        # filename = time.strftime('%Y%m%d-%H%M%S') + ".jpg"  # 文件名为当前时间
        filename = "test.jpg"
        cv2.imwrite(filename, fgmask[100:250,150:300])  # 保存位置
    elif key == 27:
        break

cap.release()
cv2.destroyAllWindows()#释放并销毁窗口

比对手势形状代码（完全行不通）
# -*- coding:utf-8 -*-

import numpy as np
import  cv2
from matplotlib import pyplot as plt

im = cv2.imread('test1.jpg')#需要识别的图
im2 = cv2.imread('ex2.jpg')#标准例图


kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(3,3))#3*3的椭圆核


img = cv2.cvtColor(im,cv2.COLOR_BGR2GRAY)#用这个方式转换的原因是最后输出时希望能看到彩色的的轮廓图
img2 = cv2.cvtColor(im2,cv2.COLOR_BGR2GRAY)

ret,thresh = cv2.threshold(img,127,255,0)
ret,thresh2 = cv2.threshold(img2,127,255,0)#将图像二值化

dst = cv2.morphologyEx(thresh,cv2.MORPH_OPEN,kernel)
dst2 = cv2.morphologyEx(thresh2,cv2.MORPH_OPEN,kernel)#将图像进行开运算

img,contours,hierarchy = cv2.findContours(dst,1,2)
max = 0
maxl = -1
for i in range(len(contours)):
    if cv2.contourArea(contours[i]) > max :
        max =  cv2.contourArea(contours[i])
        maxl = i

print(maxl)
print(len(contours))
cnt = contours[maxl]#找到需识别图中最大的轮廓，即手的轮廓

img,contours,hierarchy = cv2.findContours(dst2,1,2)
cnt2 = contours[len(contours)-1]#对照例图中手的轮廓


cv2.drawContours(im,[cnt],0,(0,255,0),3)

ret = cv2.matchShapes(cnt,cnt2,1,0.0)
print(ret)#将两轮廓进行比对

# cv2.imshow('img',img)
# cv2.imshow('thresh',thresh)
cv2.imshow('im',im)
cv2.waitKey(0)

　识别凸包代码（准确度低）

　　# -*- coding:utf-8 -*-

import numpy as np
import cv2
from matplotlib import pyplot as plt

im = cv2.imread('test.jpg')

kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(3,3))#3*3的椭圆核
img = cv2.cvtColor(im,cv2.COLOR_BGR2GRAY)#用这个方式转换的原因是最后输出时希望能看到彩色的的轮廓图

ret,thresh = cv2.threshold(img,127,255,0)

dst = cv2.morphologyEx(thresh,cv2.MORPH_OPEN,kernel)

img,contours,hierarchy = cv2.findContours(dst,1,2)
max = 0
maxl = -1
for i in range(len(contours)):
if cv2.contourArea(contours[i]) > max :
max = cv2.contourArea(contours[i])
maxl = i

print(maxl)
print(len(contours))
cnt = contours[maxl]#找到需识别图中最大的轮廓，即手的轮廓

#显示凸包
hull = cv2.convexHull(cnt,returnPoints=True)
print(hull)
cv2.polylines(im, [hull], True, (0, 255, 0), 2)

#凸性检测
# hull = cv2.convexHull(cnt,returnPoints=False)
# defect = cv2.convexityDefects(cnt,hull)
#
# for i in range(defect.shape[0]):
# s,e,f,d = defect[i,0]
# start = tuple(cnt[s][0])
# end = tuple(cnt[e][0])
# far = tuple(cnt[f][0])
# # cv2.line(im,start,end,[0,255,0],2)
# # cv2.circle(im,start,5,[0,0,255],-1)
# # cv2.circle(im,end,5,[0,0,255],-1)
# cv2.circle(im,far,5,[0,0,255],-1)

M = cv2.moments(cnt)
cx = int(M['m10']/M['m00'])
cy = int(M['m01']/M['m00'])
cv2.circle(im,(cx,cy),5,[0,0,255],-1)

cv2.imshow('img',im)
cv2.waitKey(0)

分析原因，是因为

1）背景杂质太多，且由于使用了笔记本自带摄像头，无法单独识别到手

2）光线不均匀，导致背景减除时无法准确识别到整个手的完整轮廓

综上，该方法在要求实现条件更为广泛的情况下不太行得通