【摘要】 介绍了基于机器视觉的储粮害虫智能检测系统软件部分各主要环节的具体实现。该系统运用图像差分法及自适应图像增强法提高粮虫样本图像的质量 ,利用改进的直方图阈值将粮虫从背景中分割开来 ,并运用数学形态学处理法进行了滤波。以提取出的粮虫面积、周长、复杂度为特征 ,运用基于模糊决策的分类器对粮仓中常见的 9种、7类害虫进行了分类 ,识别正确率达到 95 .2 %。更多还原

【Abstract】 An intelligent detection system of stored grain pests based on machine vision was introduced and the software realization of the main parts in the system was given. The images subtraction and self adaptive enhancement techniques were used for improving the images of stored grain pest samples, the improved histogram threshold for separating the pests from the background, and the method of mathematical morphology for filtering noise. The area perimeter and complexity of a pest were taken as its characteristics to identify the pests by using a classifier based on fuzzy decision. The correct identifications of this classifier reached 95 2%.更多还原