ass日本风韵熟妇pics男人扒开女人屁屁桶到爽|扒开胸露出奶头亲吻视频|邻居少妇的诱惑|人人妻在线播放|日日摸夜夜摸狠狠摸婷婷|制服 丝袜 人妻|激情熟妇中文字幕|看黄色欧美特一级|日本av人妻系列|高潮对白av,丰满岳妇乱熟妇之荡,日本丰满熟妇乱又伦,日韩欧美一区二区三区在线

Aiming at solving the the problems of excessive time consumption and low registration efficiency caused by the4PCS algorithm when registrating the point cloud data, a improved4PCS coarse registration method based on the3D-SIFT feature point was proposed. The point cloud data of the cherry tree was collected from four directions by DK depth camera. Firstly, a point cloud denoising framework was designed by using traight-through filtering and statistical filtering to screen high-quality three-dimensional point cloud. Secondly, the SIFT algorithm was applied to extract features from cherry tree point cloud, which reduced data dimensions and enhanced feature stability. Thirdly, the obtained set of points about source feature and target feature were used as initial data of the 4PCS algorithm, and the coarse registration was carried out. Finally, after obtaining the precise pose, the ICP algorithm was used for precision registration until the best matching state was achieved. Taking cherry tree point cloud data of different types as the experimental objects to registration experiments, the time consuming and the root maen square error indexes were introduced to evaluate the experiments. In the coarse registration stage, the results showed that the registration time of the proposed registration method was 4.16 s and 4.33 s, respectively. The root mean square error was 0.953 cm and 1.810 cm, respectively, which effectively reduced the registration error and shortened the registration time. The results of multiple precision registration experiments demonstrated that both the overall point cloud registration time and registration error achieved optimal values based on the fusion of the proposed method and the ICP algorithm in the precision registration. The whole registration time was 4.84 s and the root mean square error was 0.845 cm.