基于改进GrabCut算法的黄瓜植株图像分割

植物图像的自动分割是植物表型研究的热点问题,也是作物生长过程监测、病虫害识别等应用的核心技术之一。以黄瓜为对象,通过对图像中作物与背景特点的分析,选取EXG超绿分割和GrabCut算法进行试验研究;基于EXG超绿分割和GrabCut算法在黄瓜群体图像上的分割结果及这两种算法的优缺点,提出具有更高分割精度的改进算法。用室内室外不同生长时期的黄瓜植株图像进行试验,温室内图像和室外自然光照图像的平均分割精度分别达到96.56%和96.59%,均优于EXG超绿分割和GrabCut算法。同时表明,本文的改进算法适应性更强,具有较好的鲁棒性。

Image segmentation of cucumber plants based on improved GrabCut algorithm

The automatic segmentation technology for studying crop population canopy images could provide core algorithms for applications such as crop growth process monitoring and phenotypic parameter extraction.Based on the analysis of the cucumber plant and background features in the image,EXG super-green segmentation and GrabCut algorithm were selected for experimental research.An improved algorithm with higher segmentation accuracy was proposed based on the advantages and disadvantages of EXG super green segmentation and GrabCut algorithm.Cucumber plant images in three different growth periods of early,middle and late were tested.The performance of the algorithm was analyzed based on the segmentation accuracy ACC and algorithm processing time.The experimental results showed that the improved algorithm could get average segmentation accuracy of 96.56%on indoor images and average segmentation accuracy of 96.59%on outdoor images under natural light conditions respectively,both are better than ExG super-green segmentation and GrabCut algorithm.The experimental results also demonstrated that our improved algorithm has better adaptability and robustness.