一种基于改进非极大值抑制的农作物病斑检测方法  

A plant lesion detection method based on improved non-maxima suppression

作　　者：夏永泉[1] 黄海鹏[1] 王兵[1] XIA Yong-quan HUANG Hai-peng WANG Bing(College of Computer and Communication Engineering, Zhengzhou University of Light Industry, Z hengzhou 450001, China)

机构地区：[1]郑州轻工业学院计算机与通信工程学院,河南郑州450001

出　　处：《轻工学报》2017年第1期97-102,共6页Journal of Light Industry

基　　金：国家自然科学基金项目(61302118);河南高校青年骨干教师资助计划(2010GGJS-114)

摘　　要：针对传统边缘检测算子在植物病斑检测过程中抗噪性能差、自适应能力较弱、处理过程中采用的非极大值抑制方法容易导致伪噪声边缘出现等问题,提出了一种线性插值的非极大值抑制改进方法.该方法首先在预处理过程中使用改进的形态学滤波法替代高斯滤波,然后通过引入插值因子与邻域相关系数,将8个邻域简化为4个邻域,进行计算插值的梯度幅值与目标点的比较,替代传统方法中直接沿着梯度方向与邻域值相比较.实验结果表明:改进后的方法在抑制伪边缘点出现与适应性方面有了较好的提升,对病斑部分的分割具有非常好的有效性、准确性和鲁棒性.Since the traditional edge detection operators in a plant lesion detection process using non-maxima suppression method very easily led to spurious noise edges, poor anti-noise performance and weak adaptive capacity, a non-maxima suppression based on improved linear interpolation method was proposed. This method at first used improved morphological filtering method to replace Gaussian filtering. Then by introducing interpo- lation factor and related field coefficient,it used the target pixel four points and related fields around the gradi- ent direction coefficient linear interpolation, thus replacing the traditional method to compare with the field value directly along the gradient direction. The results showed that improved method had good improvements in terms of suppressing false edge point appears and adaptability in the lesion portion division, and had very good validity, accuracy and robustness.

关 键 词：形态学滤波 线性插值 非极大值抑制 病斑检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]