Prototype system of insects identification based on computer vision

Abstract

This paper presents the insects' image preprocessing, feature extraction and target recognition. The objective of this research was to design a new type of insect multimedia databases. Implementation of three types of basic feature extraction methods: color features of image, texture features and shape features, mainly the color vector, Gabor-wavelet transform, Fourier, GLCM, etc was done and then experimental results were analyzed. Authors suggested a new assisted GPS feature which can improve the retrieval efficiency of large insect's dataset. Designed and implemented a multi-feature asynchronous insect retrieval system, multi-feature fusion research retrieval features normalization and similarity measure algorithm. Implemented insect identification sample tested in Matlab and Visual C++ environment and the test results were analyzed. The proposed database will include audio, video and special image (such as the infrared image). This research will add an assistant feature of GPS applications, and this feature will effectively enhance the efficiency of retrieval in large dataset.