实验环境:
Python3.6+OpenCV3.4+pycharm2019
代码实现:
首先是一个
简单的不用kalman滤波的运动目标追踪代码
这里可以根据需要进行摄像头运动目标识别,只要把
camera = cv2.VideoCapture('./video/yellow_ball.mp4')
# 改成camera = cv2.VideoCapture(0) 就是摄像头内识别
完整代码:
from collections import deque
import numpy as np
import cv2
# imutils
import time
#设定阈值,HSV空间
redLower = np.array([11, 100, 100])
redUpper = np.array([20, 255, 255])
#初始化追踪点的列表
mybuffer = 64
pts = deque(maxlen=mybuffer)
#打开摄像头
camera = cv2.VideoCapture('./video/yellow_ball.mp4')
fourcc = cv2.VideoWriter_fourcc(*'XVID') # 保存文件为avi格式
fps = camera.get(cv2.CAP_PROP_FPS)
size = (int(camera.get(cv2.CAP_PROP_FRAME_WIDTH)), int(camera.get(cv2.CAP_PROP_FRAME_HEIGHT))) # 输出视频尺寸,长和宽
out = cv2.VideoWriter('new_yellow_ball.avi', fourcc, fps, size) # 输出视频
#等待两秒
time.sleep(2)
#遍历每一帧,检测红色瓶盖
while True:
#读取帧
(ret, frame) = camera.read()
if ret == False:
break
#判断是否成功打开摄像头
# if not ret:
# print('No Camera')
# break
#frame = imutils.resize(frame, width=600)
#转到HSV空间
hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
#根据阈值构建掩膜
mask = cv2.inRange(hsv, redLower, redUpper)
#腐蚀操作
mask = cv2.erode(mask, None, iterations=2)
#膨胀操作,其实先腐蚀再膨胀的效果是开运算,去除噪点
mask = cv2.dilate(mask, None, iterations=2)
#轮廓检测
cnts = cv2.findContours(mask.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[-2]
#初始化瓶盖圆形轮廓质心
center = None
#如果存在轮廓
if len(cnts) > 0:
#找到面积最大的轮廓
c = max(cnt