【python】图片标定+双目测距+SGBM+yolov8目标检测（一、图片标定）

步骤流程

1. 相机校准：在使用双目测距系统之前，需要对两个摄像头进行校准。相机校准的目的是确定两个摄像头之间的内外参数，以及消除镜头畸变等因素。
2. 使用sgbm算法创建深度图，视差图
3. 使用yolov8目标检测算法获取检测目标的中心位置坐标
4. 计算距离

相机校准

使用matlab进行校准

获得标定图片

按下按键空格保存图片，将左右两个摄像头的图片分别保存

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



camera = cv2.VideoCapture(0)

# 设置分辨率 左右摄像机同一频率，同一设备ID；左右摄像机总分辨率2560x720；分割为两个1280x720
camera.set(cv2.CAP_PROP_FRAME_WIDTH, 2560)
camera.set(cv2.CAP_PROP_FRAME_HEIGHT, 720)
num=0

while True:
    ret, frame = camera.read()
    # 裁剪坐标为[y0:y1, x0:x1] HEIGHT*WIDTH
    left_frame = frame[0:720, 0:1280]
    right_frame = frame[0:720, 1280:2560]



    cv2.imshow("left_frame ", left_frame )
    cv2.imshow("right_frame ", right_frame )
    if cv2.waitKey(1) & 0xFF == ord(' '):
        cv2.imwrite('./images/left_images/'+str(num)+'.png',img=left_frame)
        cv2.imwrite('./images/right_images/' + str(num) + '.png', img=right_frame)
        print("images_save")
        num+=1
    elif cv2.waitKey(1) & 0xFF == 27:
        break
cv2.destroyAllWindows()
camera.release()

matlab标定

标定图片效果展示

将校准后的相机内参复制到代码中

import cv2
import numpy as np


# # 加载相机参数
K_L = np.array([[724.298599316217	,0	,636.649543515655],
[0	,725.259333661568	,300.379160050881],
[0	,0	,1]])
K_R = np.array([[729.938101182192,	0,	574.732420710052],
[0	,731.188334738551,	290.944378209093],
[0	,0	,1]])

newCameraMatrix_l = np.array([[724.298599316217, 0, 640],
                            [0, 725.259333661568, 360],
                            [0, 0, 1]])
newCameraMatrix_r = np.array([[729.938101182192, 0, 640],
                            [0, 731.188334738551, 360],
                            [0, 0, 1]])

# 畸变系数,K1、K2、K3为径向畸变,P1、P2为切向畸变
dist_coeffs_L = np.array([[0.0883525071083177	,-0.103953270058449	,0.0403065973589225,0.000323856565496780,0]])
dist_coeffs_R = np.array([[0.103436996514563,	-0.181194399708597,	-7.93769896630812e-05,	0.000385968542927152	,0]])

# 设置分辨率 左右摄像机同一频率，同一设备ID；左右摄像机总分辨率2560x720；分割为两个1280x720
camera = cv2.VideoCapture(0)
camera.set(cv2.CAP_PROP_FRAME_WIDTH, 2560)
camera.set(cv2.CAP_PROP_FRAME_HEIGHT, 720)
num=0

while True:
    ret, frame = camera.read()
    # 裁剪坐标为[y0:y1, x0:x1] HEIGHT*WIDTH
    left_frame = frame[0:720, 0:1280]
    right_frame = frame[0:720, 1280:2560]

# 校正图像

    undistorted_image_L = cv2.undistort(left_frame, K_L, dist_coeffs_L,newCameraMatrix_l)
    undistorted_image_R = cv2.undistort(right_frame, K_R, dist_coeffs_R,newCameraMatrix_r)
    # 显示校正结果


    cv2.imshow('left Image', left_frame )
    cv2.imshow('right Image', right_frame )
    cv2.waitKey(1)

cv2.destroyAllWindows()