【摘要】 研究了针对空瓶特点如何设计空瓶的图像采集系统及其从空瓶图像中提取特征的方法.首先提出采用多特征综合的专家系统以适应高速检测的需要.其后又研究了基于遗传支持向量机的决策算法,提出采用遗传算法根据分类准确率来优化支持向量机的参数,以提高分类决策的性能.通过与模糊神经网络的实验比较表明,采用专家决策算法检测速度快;基于遗传支持向量机决策算法的准确率较高,达到94%以上,并具有更好的推广性能.更多还原

【Abstract】 With regard to the characteristics of empty bottle, the article focused on how to capture the images of empty bottle and extract features from the images based on machine vision. The classification-based multi-feature synthesizing expert system were put forward for high speed inspection. And this article studies the GA-SVMs(Genetic Algorithms-Support Vector Machines) to classify bottles. In GA-SVMs, the genetic algorithms were proposed to optimize the parameters of SVMs according to the accuracy rate in order to enhance the abilities of SVMs. Comparison of the advantages and disadvantages of these methods with fuzzy neural networks through experiments has shown that the speed of expert decision-making method is fast, the accuracy rate of the GA-SVM is high, at above 94%, and it has great prospects for popularity.更多还原