形状上下文在害虫识别上的应用  

Application of the Shape Context to the Insect Recognition

作　　者：邱道尹[1] 朱云[1] 

机构地区：[1]华北水利水电学院,河南郑州450011

出　　处：《华北水利水电学院学报》2011年第6期113-116,共4页North China Institute of Water Conservancy and Hydroelectric Power

摘　　要：将形状上下文引入到害虫识别中,在对害虫图像进行预处理得到二值图像的基础上,用形状上下文描述害虫的轮廓特征.首先,使用Robert算子,在二值图像中寻找害虫的轮廓并采样获取边界点集,然后利用边界点集,通过改进的对数极坐标变化计算得到目标图像的形状直方图.利用形状直方图进行2图像之间的相似性度量,计算出匹配代价.实验结果表明,跟传统的图像识别技术相比,形状上下文只要获取有限的害虫轮廓边界点,就能有效地识别出害虫.当图像发生二维变化时,改进的形状上下文算法提高了识别的准确度.In this paper,the shape context is applied to the insect recognition.The contour of insect is depicted by the shape context descriptor.Converting the insect image to binary image is preprocessing.Firstly,a set of the insect＇s contour is acquired by using the Robert operator,and then the acquired set in the modified log-polar coordinate is transformed in order to calculate the shape histogram.Finally,the cost of matching between two images＇ shape histograms is computed.According to the experiments,compared to the traditional image recognition method,the shape context can identify the insect with limited edge points of it.When the image is transformed in two-dimensional space,the improved shape context algorithm can increase the accuracy of insect recognition.

关 键 词：模式识别 形状上下文 ROBERT算子 对数极坐标 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]