基于偏好人工免疫网络多特征融合的油茶果图像识别  

Image recognition of camellia fruit based on preference for aiNET multi-features integration

作　　者：李昕[1] 李立君[1] 高自成[1] 周健[1] 闵淑辉[1] 

机构地区：[1]中南林业科技大学机电工程学院,长沙410004

出　　处：《农业工程学报》2012年第14期133-137,共5页Transactions of the Chinese Society of Agricultural Engineering

基　　金：林业公益性行业科研专项基金(201104090)

摘　　要：为提高油茶果采摘机器人机器视觉的识别率,该文提出了基于偏好人工免疫网络识别的油茶果多特征融合识别方法。在对油茶果图像进行处理的基础上,提取待识别目标区域的颜色特征、形态特征、纹理特征进行聚类,并提取典型油茶果多特征作为偏好抗体,使多特征参数在偏好免疫算法中进行有效融合。仿真试验结果表明,多特征融合的识别方法对油茶果果实的识别率在晴天时达到了90.15%,阴天时达到了93.90%。而时间复杂度基本不变,取得了较好的识别效果,该研究可为下一步油茶果采摘机器人智能采摘提供参考。To promote the recognition rates of machine-vision system in camellia fruit picking robot, the paper proposed the method of multi-feature integration using preference aiNet as the recognition algorithm. Based on the image procession of camellia fruit, the color feature, the morphology feature and the texture feature of the object region were clustered and the multi-features of camellia fruits were taken as the preference antibody, so the parameters of multi-features made effective integration in the preference aiNet. The simulation proved the accuracy of multi-feature integration reached 90.15% in the sunny day and 93.90% in the cloudy day, and the time complexity was nearly the same. This research has a certain meaning in application of forestry picking robot and provided the reference for further research of the camellia fruit picking robot.

关 键 词：机器视觉 图像处理 特征识别 偏好人工免疫网络 油茶果 采摘机器人 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]