WHY

前一阵子开始接触手势识别，觉得现在的库是真强大，13行代码就调出了手势识别，那手势识别出来后，可以做什么呢？今天，我又跟着印度小哥玩了一把隔空移物。

WHAT

什么事隔空移物呢？就是利用手指间的距离，控制图片的位置。当食指和中指并拢且在图片里面时，就更新图片的坐标（跟着食指走），所以看起来就好像被手指拖着移动了。道理很简单，我从网上找了一个透明底的苹果图片，作为素材。

HOW

要点

我用的苹果电脑，所以在读取图片路径的时候费了老大劲儿。目前只有绝对路径可以读取到，其他写法全部失败。
然后为了跟视频尺寸匹配，用了openCV的resize方法改写了图片的原始大小。

代码

from math import fabs
import cv2
import cvzone
from cvzone.HandTrackingModule import HandDetector

# read a picture which is an apple
path = '/Users/sean/Downloads/piclib/apple1.png'
imgApple = cv2.imread(path, cv2.IMREAD_UNCHANGED)

# resize the picture into proper size
widthApple, heightApple = 200, 196
apple = cv2.resize(imgApple, (widthApple, heightApple), interpolation=cv2.INTER_AREA)

# load video from camera
<参见上篇文章>

# set picture initial position
ox, oy = 50, 30

while True:
    success, img = cap.read()
    img = cv2.flip(img, 1) # flip the right and left hand
    hands, img = detector.findHands(img, flipType=False) # find hands in video

    if hands:
        lmList = hands[0]['lmList']
        # get the distance between index finger and middle finder
        length, info, img = detector.findDistance(lmList[8], lmList[12], img)
        if length < 40:
            # if the distance is lower than 40, update the picture position
            cursor = lmList[8]
            if ox < cursor[0] < ox + widthApple and oy < cursor[1] < oy + heightApple:
                ox, oy = cursor[0] - widthApple // 2, cursor[1] - heightApple // 2

    # load apple picture in the video
    img = cvzone.overlayPNG(img, apple, [ox, oy])

    cv2.imshow('Hands', img)
    cv2.waitKey(1)