基于图像增强和α角度模型的K均值小麦冠层分割算法的改进

发布时间：2019-03-14 10:02

【摘要】：[目的]本文旨在克服光照不均引起的低对比度、反光、阴影、光斑及遮挡等对大田复杂背景下小麦冠层图像分割的干扰。[方法]设计了一种结合脉冲耦合神经网络(pulse coupled neural network,PCNN)与同态滤波的自适应图像增强和基于L*a*b*颜色空间α角度模型的K均值聚类分割算法。首先,将小麦冠层图像转换到HSI颜色空间,采用自适应算法对HSI空间的I分量进行增强处理,适当调节饱和度S分量,补偿光照强度分布不均,去除阴影及拉大对比度;其次,将增强处理后的图像映射到L*a*b*颜色空间,提取a*、b*分量建立α角度模型;最后,基于α进行K均值聚类分割处理。[结果]拔节前后光照强度不一、光照不均的冬小麦冠层图像的分割试验结果表明,该算法可一定程度避免基于L*a*b*颜色空间α角度分量K均值聚类的过分割现象;改善基于HSI空间H分量K均值聚类的欠分割缺陷,且对光斑、阴影遮挡、反光突出的图像分割更完整准确。[结论]本算法可为大田复杂背景下光照多变的作物冠层图像分割提供参考方法。
[Abstract]:[aim] the aim of this paper is to overcome the interference of low contrast, reflection, shadow, spot and occlusion on wheat canopy image segmentation in complex background of field. [methods] an adaptive image enhancement algorithm based on pulse-coupled neural network (pulse coupled neural network,PCNN) and homomorphism filtering and K-means clustering algorithm based on a-angle model of Lana * color space was designed. [methods] an adaptive image enhancement algorithm based on pulse-coupled neural network and homomorphism filter was designed. Firstly, the wheat canopy image is converted to HSI color space, and the I component of HSI space is enhanced by adaptive algorithm. The S component of saturation is adjusted appropriately to compensate the uneven distribution of light intensity, remove the shadow and enlarge the contrast. Secondly, the enhanced image is mapped to the color space, and the 伪-angle model is built. Finally, the K-means clustering segmentation is carried out based on 伪. [results] the experimental results of segmentation of winter wheat canopy images with different intensity and uneven illumination before and after jointing show that the proposed algorithm can avoid the over-cut of K-means clustering of 伪-angle component in color space. It improves the defect of under-segmentation based on K-means clustering of H-component in HSI space, and it is more complete and accurate for image segmentation of speckle, shadow occlusion and reflective highlight. [conclusion] this algorithm can provide a reference method for crop canopy image segmentation under complex background.
【作者单位】：南京农业大学国家信息农业工程技术中心南京农业大学信息科学与技术学院中国移动通信集团浙江有限公司嘉兴分公司
【基金】：国家重点研发计划项目(2016YFD0300607) 江苏省农业科技自主创新资金项目[CX(14)2116]
【分类号】：S512.1;TP391.41

【相似文献】