Insect Identification Based On Image Pattern Recognition Technology Research

This paper studied the images of Lepidopteral samples with technique of Digital Image Process and Fuzzy Model Identify, achieved the aim of auto-identify similar species. The main content and achievement as follow:(1) Design of Lepidopter Identify System. This paper expatiated on constructing image acquisition device and the flow of digital image processing and identify.(2) Digital image pre-process. This paper introduced the techniques of smooth and noise-suppress, adaptive threshold segment, mathematical Morphology and image segment, and introduced the effect and result using pre-process techniques at Lepidopter. Found a method to segment antenna well while traditional way can't satisfy.(3) Characters distilling. This paper distilled some global characters like perimeter, area, form parameter, vertical axis, horizontal axis, eccentricity, roundness and some local characters like light area's number, ventral length, ventral width, ventral ratio of Lepidopter.(4) Characters selecting, classify and identify. This paper introduced theory of Fuzzy Sets briefly, realized characters selecting with fuzzy c-mean cluster, designed primary classifier based Lepidopteral light area's number, designed second classifier with Fuzzy Model Identify based maximal subjection principle, reduced the procedure time.(5) Software realizing. At end, this paper introduced the thinking and method of software part. Programmed the software part mix Visual C++ and Basic because of VC's quick and VB's easy.