Identification Algorithm of Winged Insects Based on Hybrid Moment Invariants

Automated species identification based on image processing and pattern recognition, as a small but growing domain, has been actively researched in recent years. Feature extraction is always the key step. This paper proposes a hybrid moment invariants method to extract the features of image. The main idea of this method is to combine Hu's invariants and blur- affine moment invariants to identify winged insects obtained under unknown illumination and blurring conditions and it is exemplified by left forewing images of dragonflies. The new method has more advantages such as insensitive to translation, scale change, rotation, as well as affine transform and blurring. The experimental results on a small set of samples showed that the method can achieve higher identification rate.