CAJ | 학술논문

提出一种基于双目立体视觉的复杂背景下的牛体点云获取方法。在点云获取过程中，针对牛不是保持不动的实际情况，采用基于被动视觉技术的双目立体视觉方法；在相机标定过程中，进行单目标定获取相机内部参数，统一标定获取相机外部参数，针对牛体毛色及所处环境复杂的情况，采用基于贝叶斯的皮肤检测算法提取牛体图像，采用基于SIFT的特征点提取和匹配方法实现特征点的匹配，利用计算机视觉中的极线几何原理，剔除误匹配点，通过相机的成像模型求取牛体的三维点云。实验验证了该方法能够在复杂背景下获取牛体点云，并得到较好的点云数据，较好解决了双目立体视觉中相机标定和立体匹配两个较关键和困难的问题。
제출일충기우쌍목입체시각적복잡배경하적우체점운획취방법。재점운획취과정중，침대우불시보지불동적실제정황，채용기우피동시각기술적쌍목입체시각방법；재상궤표정과정중，진행단목표정획취상궤내부삼수，통일표정획취상궤외부삼수，침대우체모색급소처배경복잡적정황，채용기우패협사적피부검측산법제취우체도상，채용기우SIFT적특정점제취화필배방법실현특정점적필배，이용계산궤시각중적겁선궤하원리，척제오필배점，통과상궤적성상모형구취우체적삼유점운。실험험증료해방법능구재복잡배경하획취우체점운，병득도교호적점운수거，교호해결료쌍목입체시각중상궤표정화입체필배량개교관건화곤난적문제。
To obtain the point cloud data of cattle in a complex situation,a method based on binocular stereo vision was presen-ted.Considering that the cattle do not always stand motionless during the process of obtaining point cloud,the binocular stereo vision of passive vision technology was adopted.During camera calibration,single camera calibration was conducted firstly,and then extrinsic parameters were obtained through calibrating the two cameras at the same time.To get the cattle from the com-plex situation,a skin detection algorithm based on Bayesian classification was implemented.Feature points extraction and matc-hing method based on SIFT algorithm was employed and epipolar geometric principle was utilized to eradicate the false matching points.Finally,the three-dimensional coordinate was gotten by means of imaging model of camera.The experimental results show that the proposed method is feasible in the complex situation,and it can obtain good point cloud data.The two critical and difficult issues of camera calibration and stereo matching in binocular stereo are solved availably.