棉花铺膜播种机导航路线图像检测方法

研究了棉花铺膜播种机田间作业时导航路线和田端的图像检测算法。针对自然环境下的棉花播种作业图像,采用Daubechies小波变换对处理区域进行平滑滤波;针对第1帧图像,寻找图像处理区域的垂直累计直方图的波谷,以此为基础,通过寻找局部窗口累计直方图波谷的方法,从图像底端逐行向上寻找各行候补点;对于非第1帧图像,采用当前帧与前帧导航路线相关联的方法分段寻找候补点群;最后基于过已知点Hough变换拟合出导航路线。实验证明,采用的算法可以快速、准确地检测出棉花铺膜播种作业时的导航路线及棉田田端,平均每帧图像处理时间为72.02 ms,满足铺膜播种机实际播种作业的需求。

Image Detection Method of Navigation Route of Cotton Plastic Film Mulch Planter

An image algorithm to detect the navigation route and the filed end while the cotton plastic film mulch planter was working in the field was studied. The method of Daubechies wavelet transform was used to smooth and filter the image of the cotton sowing operation under the natural environment. The trough of the vertical cumulative histogram in the image processing region was detected in the first frame image. Based on this, the candidate points were detected scanning from the bottom row of the image to the top by using the method of detecting the trough of the local window cumulative histogram. For other frame images, the candidate points were detected segment by segment by the association between the navigation route of the current frame and that of the previous frame. Then the navigation route was fitted based on the passing a known point Hough transform (PKPHT). The experiments showed that the algorithm could quickly and accurately detect the navigation route and the field end while the cotton plastic film mulch sowing work. The average processing time of each frame was 72.02ms, which could meet the actual sowing work requirement of the film mulch planter.