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

Crop height and cut-edge are important factors to be considered in unmanned rice wheat combine harvester, because the height of sowing wheel is adjusted according to the crop height and cut-edge provides navigation information. Therefore, field crop height and cut-edge information were extracted based on binocular vision. The 3D data and RGB image were acquired by binocular camera. The 3D data on the flat ground were filtered by voxels and through filters, and the filtered data was fitted to the initial plane by RANSAC algorithm. The real-time plane of harvesting operation was calculated with posture changes of harvester reflected by IMU data, and the 3D data was transformed into the actual height according to the distance from point to plane. An improved method combining density peak clustering and K-means clustering was proposed to classify the height data. At the same time, the RGB image was normalized and then segmented by Otsu algorithm to extract the upper region of crop. The common region of the cluster with the largest cluster center value and the upper crop region were obtained, and the mean value of the height data belonging to the common region was calculated to obtain the crop height. Based on the cross-correlation between the height data series and the model function, the cut-edge points were extracted. The cut-edge points were fitted to the cut-edge line by the least square method. According to the current boundary line, the candidate range of the next frame data cut-edge points was predicted. The heading deviation and lateral deviation were calculated by the cut-edge line. Experiments showed that this method could effectively extract the crop height and cut-edge information, and the mean absolute error of height was 0.043m and the correct rate of boundary recognition was 93.30% under the complex harvest scenes including sparse, missed cutting and rutting. The average angle error of heading deviation was 1.04°, and the average absolute error of lateral deviation was 0.084m. Therefore, the method had application value to unmanned self-adaptive control of combine harvester.