基于mediapipe和opencv的手势识别

效果图如下

import mediapipe as mp
import cv2
import numpy as np
#定义获取指关节角度的函数
def get_angle(v1,v2):
    angle=np.dot(v1,v2)/(np.sqrt(np.sum(v1*v1))*np.sqrt(np.sum(v2*v2)))
    angle=np.arccos(angle)/3.14*180
    return angle

#定义获取手势的函数
def get_str_guester(up_fingers,list_point):
    if len(up_fingers)==1 and up_fingers[0]==8:  #当只有手指8竖起时
        v1=list_point[6]-list_point[7]
        v2=list_point[8]-list_point[7]
        #通过判断两个关节的角度来判断该手指是否是弯曲的
        angle=get_angle(v1,v2)
        if angle<160:
            str_guester="9"
        else:
            str_guester="1"
    elif len(up_fingers)==1 and up_fingers[0]==4:
        str_guester="Good"
    elif len(up_fingers)==1 and up_fingers[0]==20:
        str_guester="Bad"
    elif len(up_fingers)==2 and up_fingers[0]==8 and up_fingers[1]==12:
        str_guester="2"
    elif len(up_fingers)==2 and up_fingers[0]==4 and up_fingers[1]==20:
        str_guester="6"
    elif len(up_fingers)==2 and up_fingers[0]==4 and up_fingers[1]==8:
        str_guester="8"
    elif len(up_fingers)==3 and up_fingers[0]==8 and up_fingers[1]==12 and up_fingers[2]==16:
        str_guester="3"
    elif len(up_fingers) == 3 and up_fingers[0] == 4 and up_fingers[1] == 8 and up_fingers[2] == 12:
        str_guester="7"
    elif len(up_fingers) == 3 and up_fingers[0] == 4 and up_fingers[1] == 8 and up_fingers[2] == 20:
        str_guester = "Rock"
    elif len(up_fingers) == 4 and up_fingers[0] == 8 and up_fingers[1] == 12 and up_fingers[2] == 16 and up_fingers[3] == 20:
        str_guester="4"
    elif len(up_fingers) == 5 :
        str_guester = "5"
    elif len(up_fingers) == 0:
        str_guester = "10"
    else:
        str_guester=" "
    return str_guester


if __name__=="__main__":
    cap=cv2.VideoCapture(0);
    mpHands=mp.solutions.hands
    hands=mpHands.Hands()
    mpDraw=mp.solutions.drawing_utils
    while True:
        success,img=cap.read()#读取每一张图片
        if not success:
            continue
        image_height,image_width,imagedepth=np.shape(img)
        imgRGB=cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
        results=hands.process(imgRGB)#将图片带入到函数中处理，得到result
        if results.multi_hand_landmarks:
            hand=results.multi_hand_landmarks[0]   #找到第一个手并画出来
            mpDraw.draw_landmarks(img,hand,mpHands.HAND_CONNECTIONS)
            list_point=[] #放置关键点的坐标list
            for i in range(21):
                pos_x=hand.landmark[i].x*image_width
                pos_y=hand.landmark[i].y*image_height
                list_point.append( [int(pos_x),int(pos_y)] )
            #构造凸包
            list_point=np.array(list_point,dtype=np.int32)
            hull_index=[0,1,2,3,6,10,14,19,18,17,10]
            hull=cv2.convexHull(list_point[hull_index,:])
            cv2.polylines(img,[hull],True,(0,255,0),2) #绘制凸包
            #查找哪些手指指尖在手掌外
            n_fig=-1
            obj_finger=[4,8,12,16,20]
            up_fingers=[] #存放结果
            for i in obj_finger:
                pt=(int(list_point[i][0]) , int(list_point[i][1]))
                dist=cv2.pointPolygonTest(hull,pt,True)  #判断这个点是否在凸包里面
                if dist<0:
                    up_fingers.append(i)
            #根据得到的手指判断手势
            str_guester=get_str_guester(up_fingers,list_point)
            cv2.putText(img,'%s'%(str_guester),(90,90),cv2.FONT_HERSHEY_SIMPLEX,3,(255,255,0),4,cv2.LINE_AA)  #在视频中显示出来判断的手势
            #for i in obj_finger:
                #pos_x=hand.landmark[i].x*image_width
                #pos_y=hand.landmark[i].y*image_height
                #cv2.circle(img,(int(pos_x,pos_y)),3,(0,255,255),-1)
        cv2.imshow("hand",img)#显示结果
        key=cv2.waitKey(1)&0xFF
        if key==ord('q'):
            break
    cap.release()

#代码来源网络，如有侵权，请联系作者#