论文部分内容阅读
在执行机构对目标物体进行自主定位过程中,定位误差的实时计算、误差修正和状态分析往往比较困难.为此,提出基于三帧差法的Kalman滤波算法进行末端动态捕捉,利用反向传输(BP)神经网络分类思想进行目标识别,基于点云库的点云提取和处理算法,获得末端和目标物体的空间坐标.最后,将散乱点群进行网格化和3D空间插值.实验结果表明,算法能实时检测并跟踪运动末端,预测精度达到99%,且目标物体的识别率为99%,并可在短时间内修正定位误差,使末端中心点逐步收敛到目标质心,自主定位成功.用三维拟合法对算法的有效性进行验证,并对定位过程进行了状态分析.新算法能完成执行机构的自主定位,省去了相机标定过程,提高了系统效率.
It is often difficult to perform real-time calculation, error correction and state analysis of the positioning error in executing the autonomous positioning of the target object.For this purpose, a Kalman filtering algorithm based on the three-frame difference method is proposed to carry out dynamic acquisition at the end and use reverse transmission BP neural network classification ideology, points cloud library based on point cloud extraction and processing algorithms to obtain the spatial coordinates of the end and the target object.Finally, the scattered point group grid and 3D spatial interpolation.The experimental results show that, The algorithm can detect and track the end of motion in real time, the prediction accuracy reaches 99%, and the recognition rate of the target object is 99%, and the positioning error can be corrected in a short time so that the end center point gradually converges to the target centroid and the autonomous positioning is successful. The validity of the algorithm is verified by three-dimensional fitting method, and the state analysis of the positioning process is carried out.The new algorithm can complete the autonomous positioning of the actuator, eliminating the camera calibration process and improving the system efficiency.