基于凹点检测的玉米穗粘连图像分割算法

Research on corn ears adhesion image segmentation algorithm based on concave point detection

为了精确提取玉米考种参数,实现多种玉米穗粘连的图像分割,提出一种基于凹点检测的玉米穗粘连图像的分割算法。首先,提取原始RGB彩色图像的红色分量图,经阈值分割、开操作等预处理,对图像进行降噪、去毛刺,通过轮廓查找依次获取玉米穗的连通域、绘制凸包以及凹区域,在设置的距离阈值范围内,迭代计算两两凹区域轮廓的最小欧氏距离,并进行凹点匹配,完成一次图像分割;然后,对依然粘连的玉米穗图像进行二次分割,根据检测到的凹区域个数,对连通域进行分割处理。实验结果表明,相较于其他分割算法,提出算法降低了欠分割率,提高了分割准确率,欠分割率仅为4.8%,分割准确率达到92.9%。

In order to accurately extract the phenotypic parameters of corns and realize the image segmentation of multiple corn ears adhesion,a segmentation algorithm of corn ear adhesion images based on pit detection was proposed.Firstly,the red component map of the original RGB color image was extracted.After preprocessing such as threshold segmentation and opening operation,the image was denoised and deburred.Through contour searching,the connected region of corn ear,convex hull and concave region were obtained in turn.Within the set distance threshold range,the minimum Euclidean distance of the contour of each two concave areas was iteratively calculated,and the concave points were matched to complete the first image segmentation.Then,secondary segmentation was performed on the still adhesion corn ear image.According to the different numbers of detected concave areas,the connected areas were segmented differently.Experimental results show that compared with other segmentation algorithms,the proposed algorithm reduces the under-segmentation rate and improves the segmentation accuracy.The under-segmentation rate is only 4.8%,and the segmentation accuracy rate reaches 92.9%.