參考鏈接：雙目視覺 標定 （基於MATLAB）_一顆小樹x的博客-CSDN博客_matlab雙目校正
自制的標定數據集，必須用自己相機拍攝照片制作數據集
標定板下載：pattern.png (1830×1330) (opencv.org)
import cv2
import numpy as np
# 左鏡頭的內參，如焦距
left_camera_matrix = np.array([[516.5066236,-1.444673028,320.2950423],[0,516.5816117,270.7881873],[0.,0.,1.]])
right_camera_matrix = np.array([[511.8428182,1.295112628,317.310253],[0,513.0748795,269.5885026],[0.,0.,1.]])
# 畸變系數,K1、K2、K3為徑向畸變,P1、P2為切向畸變
left_distortion = np.array([[-0.046645194,0.077595167, 0.012476819,-0.000711358,0]])
right_distortion = np.array([[-0.061588946,0.122384376,0.011081232,-0.000750439,0]])
# 旋轉矩陣
R = np.array([[0.999911333,-0.004351508,0.012585312],
[0.004184066,0.999902792,0.013300386],
[-0.012641965,-0.013246549,0.999832341]])
# 平移矩陣
T = np.array([-120.3559901,-0.188953775,-0.662073075])
size = (640, 480)
R1, R2, P1, P2, Q, validPixROI1, validPixROI2 = cv2.stereoRectify(left_camera_matrix, left_distortion,
right_camera_matrix, right_distortion, size, R,
T)
# 校正查找映射表,將原始圖像和校正後的圖像上的點一一對應起來
left_map1, left_map2 = cv2.initUndistortRectifyMap(left_camera_matrix, left_distortion, R1, P1, size, cv2.CV_16SC2)
right_map1, right_map2 = cv2.initUndistortRectifyMap(right_camera_matrix, right_distortion, R2, P2, size, cv2.CV_16SC2)
print(Q)
cv2.stereoRectify（）函數
- 示例：
R1, R2, P1, P2, Q, validPixROI1, validPixROI2 = cv2.stereoRectify(left_camera_matrix, left_distortion,right_camera_matrix, right_distortion, size, R, T)- 作用：為每個攝像頭計算立體校正的映射矩陣R1, R2, P1, P2
- 參數：
- left_camera_matrix：左相機內參
- left_distortion：左相機畸變系數
- right_camera_matrix：右相機內參
- right_distortion：右相機畸變系數
- size：單邊相機的圖片分辨率
- R：旋轉矩陣
- T：平移矩陣
- 返回值：
- R1, R2：R1-輸出矩陣，第一個攝像機的校正變換矩陣（旋轉變換）；R2-輸出矩陣，第二個攝像機的校正變換矩陣（旋轉變換）
- P1, P2：P1-輸出矩陣，第一個攝像機在新坐標系下的投影矩陣；P2-輸出矩陣，第二個攝像機在新坐標系下的投影矩陣
# -*- coding: utf-8 -*-
import numpy as np
import cv2
# import camera_configs
import random
import math
WIN_NAME = 'Deep disp'
cv2.namedWindow(WIN_NAME, cv2.WINDOW_AUTOSIZE)
frame = cv2.imread("yojuh-wewwn-004.jpg")
img_color = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
frame1 = frame[0:480, 0:640]
frame2 = frame[0:480, 640:1280] # 割開雙目圖像
imgL = cv2.cvtColor(frame1, cv2.COLOR_BGR2GRAY) # 將BGR格式轉換成灰度圖片
imgR = cv2.cvtColor(frame2, cv2.COLOR_BGR2GRAY)
# cv2.remap 重映射，就是把一幅圖像中某位置的像素放置到另一個圖片指定位置的過程。
# 依據MATLAB測量數據重建無畸變圖片
img1_rectified = cv2.remap(imgL, left_map1, left_map2, cv2.INTER_LINEAR)
img2_rectified = cv2.remap(imgR, right_map1, right_map2, cv2.INTER_LINEAR)
imageL = cv2.cvtColor(img1_rectified, cv2.COLOR_GRAY2BGR)
imageR = cv2.cvtColor(img2_rectified, cv2.COLOR_GRAY2BGR)
# SGBM-室外
blockSize = 8
img_channels = 3
stereo = cv2.StereoSGBM_create(minDisparity=1,
numDisparities=64,
blockSize=blockSize,
P1=8 * img_channels * blockSize * blockSize,
P2=32 * img_channels * blockSize * blockSize,
disp12MaxDiff=-1,
preFilterCap=1,
uniquenessRatio=10,
speckleWindowSize=100,
speckleRange=100,
mode=cv2.STEREO_SGBM_MODE_HH)
disparity = stereo.compute(img1_rectified, img2_rectified) # 計算視差
disp = cv2.normalize(disparity, disparity, alpha=0, beta=255, norm_type=cv2.NORM_MINMAX, dtype=cv2.CV_8U) # 歸一化函數算法
dis_color = disparity
dis_color = cv2.normalize(dis_color, None, alpha=0, beta=255, norm_type=cv2.NORM_MINMAX, dtype=cv2.CV_8U)
dis_color = cv2.applyColorMap(dis_color, 2)
cv2.imshow("depth", dis_color)
cv2.imshow("left", frame1)
cv2.imshow(WIN_NAME, disp) # 顯示深度圖的雙目畫面
cv2.waitKey()
# 銷毀內存
cv2.destroyAllWindows()
img1_rectified = cv2.remap(imgL, left_map1, left_map2, cv2.INTER_LINEAR)：重映射，即把一幅圖像內的像素點放置到另外一幅圖像內的指定位置，俗稱“拼接”
我們可以通過cv.remap()函數來將img2映射到img1對應位置上並合成
cv2.StereoSGBM_create()函數為opencv集成的算法；我們只需關注blockSize。 使用方法為：
Safe-Bike/sgbm_img.py at main · yzfzzz/Safe-Bike (github.com)
