基于HSV空间再生稻植株与土壤背景图像分割

Image Segmentation for Identifying Ratooning Rice Area from Soil Background Based on HSV Space

针对再生稻收割机视觉导航的稻田图像分割问题,结合再生稻植株的生长特点和再生稻避莊的要求,利用相机于农田采集再生稻图片,结合RGB、HSV、YCr Cb空间中的常用灰度化因子,进行灰度化对比试验并分析其直方图特征,得出在HSV空间的S分量灰度化;采用最大类间方差法(Otsu)得到初步分割阈值T,经进一步分析为保留较完整的不同成熟度再生稻植株特征,加入修正因子-a得到阈值T-a对图像二值化;再通过数学形态学,面积法过滤等后续处理,形成收割机行走的左右边界区域。结果表明:处理1副像素419×310的图像平均耗时0.053 s,可满足今后的实时性要求,分割出的图像基本上反应了再生稻植株的走势特征,与人眼判断植株边缘位置基本相符合。

For the issue of image segmentation in the ratooning rice field to the vision navigation of the harvester , com-bined with the growth characteristics of the plant and the requirement to avoid ratooning rice , a method to the image was proposed .Taking the picture from field environment , combained with the commonly gray scale factor in the RGB、HSV、YCrCb color space , the contrast testing about graying image was took , and the histogram features was analysised .It was varianted in S characteristic by HSV space , and combined with the Otsu to get the initial segmentation threshold T .In or-der to maintain the integrity of the characteristics in the ratooning rice area with different maturity , the modified factor -awas added to get the segmentation threshold T-a.Then the graying image was binarized by the modified threshold , and used the mathematical morphology , filtered the small area with other subsequent process .Finally , the left and right boundary region to the harvester was formed .The results demonstrate that the average time of processing a 419 ＆#215;310 pixel image was 0 .053 s and met the need of real-time in the future .The segmented image basically reflected the trend charac-teristics of the ratooning rice area , which was consistent with the human vision to identify the edge position of the plant .