基于四维向量的频域匹配法实现工件位姿识别

Frequency Domain Matching Method Based on Four-Dimensional Vector to Realize Workpiece Pose Recognition

针对工业环境下识别无序摆放工件时存在的识别准确率低与配准时间长等问题,提出一种频域的四维向量匹配方法(FDVM)识别无序摆放工件的位姿。首先,用霍夫变换提取图像频谱图的特征直线,并提出最邻像素投影点插值方法获取该特征直线上的特征点集;其次,以该特征点集的灰度强度及灰度变化梯度构建四维向量,形成基于四维向量的旋转角匹配方法;用该方法实现工件的模板图像与无序摆放工件的特征配准,最终获得待配准工件的旋转角度和位置信息。实验结果表明,相较于SIFT法和SURF法,FDVM法计算时间分别降低89.3%和83.9%,角度匹配误差分别下降54.7%和53.6%;FDVM法能有效计算出工件旋转角度和位置,对无序摆放工件可以提升配准效率且实现精确配准。

To solve the problems of low recognition accuracy and long registration time in complex industrial conditions for disorderly placed workpieces,a novel method of posture recognition,frequency domain four-dimensional vector matching(FDVM),is proposed.Image’s characteristic lines of the image spectrogram were extracted using Hough transform,firstly.And then the nearest pixel projection point interpolation method was proposed to obtain feature point set of these characteristic lines.Secondly,four-dimensional vectors were constructed based on the gray intensity and gray change gradient of these feature point set.thus,a rotation angle matching method is constructed with the four-dimensional vectors.The experimental result shows that the calculation time of FDVM method was 89.3%and 83.9%lower than that of SIFT method and SURF method,and its angle matching error decrease 54.7%and 53.6%compared with both methods,respectively.These results indicate that the proposed method could effectively calculate the rotation angle and position of the workpiece.The FDVM can improve the registration efficiency and achieve accurate registration for disorderly placement of workpieces.A processing strategy with practical application value was proposed for real-time workpiece recognition under complex working conditions.