一种简单可靠的昆虫扑翼运动图像序列自动分析方法

摘要

Nowadays, the high-speed camera and multi-camera setup are widely used to record sequences of insects' erratic motion, however the methods for analysis these images sequences are more or less unsatisfactory.Here a simple but robust method for this automatic analysis of insect flight had been given and examined.In the sequences, the difference between adjacent frames is fully utilized, and then the feature points (tip, base, head, and abdomen) and contours (leading edge, trailing edge and body) are calculated out.The method was realized in MATIAB and tested under motion sequences of virtual insects and real insects.The performance showed that this method is simple but robust, no need of any manual intervention, unspecific insects, small calculating amount , good results even when the wings and body partly overlap or wings have obvious transparency.The method provides a practical way to solve morphology features extraction problem of insect flight images and it greatly save the time of insect flight images analysis.%目前高速摄影相机和多相机系统广泛地用于记录昆虫奇妙的飞行运动,然而分析这些运动图像序列的方法却不能让人满意.提出并测试了一种简单可靠的昆虫飞行图像的自动分析方法.该方法充分地利用了图像序列中相邻两帧间的差别,可自动提取出昆虫的特征点(翼尖点,翼根点,头部端点,腹部端点)和轮廓线(翼前缘轮廓,翼后缘轮廓).采用MATLAB实现并使用虚拟昆虫和真实昆虫运动图像序列测试了该方法.测试结果表明,该方法简单可靠,无需人工干预,不针对特定昆虫,运算量也较小,且能够提供较好的识别结果,甚至在处理昆虫的身体和翅膀部分重叠或者昆虫的翅膀图像呈现明显的透明的图像序列时候结果依然不错.该方法给昆虫图像的形态学特征自动识别难题提供了一个切实可行的解决方法,节省了昆虫扑翼运动图像的分析时间.