cmvr-es/python/hand_eye_calibration/compute_in_hand.py

# coding=utf-8

"""
眼在手上 用采集到的图片信息和机械臂位姿信息计算 相机坐标系相对于机械臂末端坐标系的 旋转矩阵和平移向量
A2^{-1}*A1*X=X*B2*B1^{−1}
"""
import os
import logging


import  yaml

import cv2
import numpy as np
from scipy.spatial.transform import Rotation as R


from libs.auxiliary import find_latest_data_folder
from libs.log_setting import CommonLog

from save_poses import poses_main

np.set_printoptions(precision=8,suppress=True)

logger_ = logging.getLogger(__name__)
logger_ = CommonLog(logger_)


current_path = os.path.join(os.path.dirname(os.path.abspath(__file__)),"eye_hand_data")

images_path = os.path.join("eye_hand_data",find_latest_data_folder(current_path))
file_path = os.path.join(images_path,"RobotToolPose.csv")  #采集标定板图片时对应的机械臂末端的齐次变换矩阵 从 第一行到最后一行 需要和采集的标定板的图片顺序进行对应


with open("config.yaml", 'r', encoding='utf-8') as file:
    data = yaml.safe_load(file)

XX = data.get("checkerboard_args").get("XX") #标定板的中长度对应的角点的个数
YY = data.get("checkerboard_args").get("YY") #标定板的中宽度对应的角点的个数
L = data.get("checkerboard_args").get("L")   #标定板一格的长度  单位为米


def func():


    # 设置寻找亚像素角点的参数，采用的停止准则是最大循环次数30和最大误差容限0.001
    criteria = (cv2.TERM_CRITERIA_MAX_ITER | cv2.TERM_CRITERIA_EPS, 30, 0.001)

    # 获取标定板角点的位置
    objp = np.zeros((XX * YY, 3), np.float32)

    objp[:, :2] = np.mgrid[0:XX, 0:YY].T.reshape(-1, 2)     # 将世界坐标系建在标定板上，所有点的Z坐标全部为0，所以只需要赋值x和y
    objp = L*objp


    obj_points = []     # 存储3D点
    img_points = []     # 存储2D点


    images_num = [f for f in os.listdir(images_path) if f.endswith('.jpg')]

    for i in range(1, len(images_num) + 1):   #标定好的图片在images_path路径下，从0.jpg到x.jpg

        image_file = os.path.join(images_path,f"{i}.jpg")

        if os.path.exists(image_file):

            logger_.info(f'读 {image_file}')


            img = cv2.imread(image_file)
            gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

            size = gray.shape[::-1]
            ret, corners = cv2.findChessboardCorners(gray, (XX, YY), None)

            if ret:


                obj_points.append(objp)

                corners2 = cv2.cornerSubPix(gray, corners, (5, 5), (-1, -1), criteria)  # 在原角点的基础上寻找亚像素角点
                if [corners2]:
                    img_points.append(corners2)
                else:
                    img_points.append(corners)

                    # 绘制角点并保存图片
                cv2.drawChessboardCorners(img, (XX, YY), corners2 if corners2 is not None else corners, ret)
                corner_folder = os.path.join(images_path, "corner")
                os.makedirs(corner_folder, exist_ok=True)  # 如果不存在就创建
                save_path = os.path.join(corner_folder, f"corner_{i}.jpg")
                cv2.imwrite(save_path, img)
                logger_.info(f"保存带角点的图片到: {save_path}")

    N = len(img_points)

    # 标定,得到图案在相机坐标系下的位姿
    ret, mtx, dist, rvecs, tvecs = cv2.calibrateCamera(obj_points, img_points, size, None, None)

    # logger_.info(f"内参矩阵:\n:{mtx}" ) # 内参数矩阵
    # logger_.info(f"畸变系数:\n:{dist}")  # 畸变系数   distortion cofficients = (k_1,k_2,p_1,p_2,k_3)

    print("-----------------------------------------------------")


    tool_pose = np.loadtxt(file_path,delimiter=',')

    R_tool = []
    t_tool = []

    N = tool_pose.shape[0] // 4  # 矩阵个数，每个矩阵占4行

    for i in range(N):
        mat = tool_pose[4*i:4*i+4, :]  # 取第i个4x4矩阵
        R_tool.append(mat[0:3, 0:3])   # 提取旋转矩阵
        t_tool.append(mat[0:3, 3])     # 提取平移向量

    R, t = cv2.calibrateHandEye(R_tool, t_tool, rvecs, tvecs, cv2.CALIB_HAND_EYE_TSAI)



    return R,t

if __name__ == '__main__':

    # 旋转矩阵
    rotation_matrix, translation_vector = func()

    # 将旋转矩阵转换为四元数
    rotation = R.from_matrix(rotation_matrix)
    quaternion = rotation.as_quat()
    x, y, z = translation_vector.flatten()

    logger_.info(f"旋转矩阵是:\n {            rotation_matrix}")

    logger_.info(f"平移向量是:\n {            translation_vector}")

    logger_.info(f"四元数是：\n {             quaternion}")