Auto-classification of insect images based on color histogram and GLCM

Aims to provide general technicians who manage pects in production with a convenient way to recognize them, a novel method to classify insects by analyzing color histogram and GLCM (Gray-Level Co-occurrence Matrices) of wing images is proposed. The wing image of lepidopteran insect is preprocessed to get the ROI (Region of Interest); then the color image is first converted from RGB(Red-Green-Blue) to HSV (Hue-Saturation-Value) space, and the 1D color histograms of ROI are generated from hue and saturation distributions. Afterward, the color image is converted to grayscale one, rotated and transformed to a standard position, and their GLCM features are extracted. Matching is first undergone by computing the correlation of the histogram vectors between testing and template images; if the correlation is higher than certain threshold, then their GLCM features are further matched. The winner-take-all policy is adopted in deciding most matched species in k nearest neighbors. The method is tested at the lepidopteran insect database with 100 species. The recognition rate is as high as 71.1%. An ideal time performance is also achieved. The experimental results testify the efficiency of proposed method.