1 解释说明
单目相机测距常用的方法就是相似三角形法。
注意:1in = 25.4mm
举个例子,假设在离相机距离 D = 45cm = (450/25.4=17.7in)的地方放一张标准的8.27in x 11.69英寸W = 11.69(23cm*16cm)的A4纸并且拍下一张照片。测量出照片中A4纸的像素宽度为 P = 874 像素。因此焦距 F 是:
F = (874px x 17.7in) /11.69in = 1323
当相机移动靠近或者离远物体或者目标时,可以用相似三角形来计算出物体离相机的距离:
D’ = (W x F) / P
2 代码
import numpy as np
import cv2
KNOWN_DISTANCE = 17.7
KNOWN_WIDTH = 11.69
KNOWN_HEIGHT = 8.27
# 定义目标函数
def find_marker(image):
gray_img = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) # 将彩色图转化为灰度图
gray_img = cv2.GaussianBlur(gray_img, (5, 5), 0) # 高斯平滑去噪
edged_img = cv2.Canny(gray_img, 35, 125) # Canny算子阈值化
# 获取纸张的轮廓数据
img, countours, hierarchy = cv2.findContours(edged_img.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
# print(len(countours))
# 获取最大面积对应的点集
c = max(countours, key=cv2.contourArea)
# 最小外接矩形
rect = cv2.minAreaRect(c)
return rect
# 定义距离函数
def distance_to_camera(knownWidth, focalLength, perWidth):
return (knownWidth * focalLength) / perWidth
# 计算摄像头的焦距(内参)
def calculate_focalDistance(img_path):
first_image = cv2.imread(img_path)
# cv2.imshow('first image', first_image)
# 获取矩形的中心点坐标,长度,宽度和旋转角度
marker = find_marker(first_image)
print("org图片中A4纸的宽度:f%", marker[1][0])
focalLength = (marker[1][0] * KNOWN_DISTANCE) / KNOWN_WIDTH
print('焦距 = ', focalLength)
return focalLength
# 计算摄像头到物体的距离
def calculate_Distance(image_path, focalLength_value):
image = cv2.imread(image_path)
# 获取矩形的中心点坐标,长度,宽度和旋转角度, marke[1][0]代表宽度
marker = find_marker(image)
distance_inches = distance_to_camera(KNOWN_WIDTH, focalLength_value, marker[1][0])
box = cv2.boxPoints(marker)
# print("Box = ", box)
box = np.int0(box)
# print("Box = ", box)
cv2.namedWindow('img', cv2.WINDOW_NORMAL)
cv2.drawContours(image, [box], -1, (0, 255, 0), 2)
cv2.putText(image, "%.2fcm" % (distance_inches * 2.54), (image.shape[1] - 1000, image.shape[0] - 100),
cv2.FONT_HERSHEY_SIMPLEX, 2.0, (0, 255, 0), 3)
cv2.imshow("img", image)
if __name__ == "__main__":
img_path = "org.jpg"
focalLength = calculate_focalDistance(img_path)
calculate_Distance("test.jpg", focalLength)
cv2.waitKey(0)
cv2.destroyAllWindows()
pass
org.jpg
test.jpg
结果图: