Performance Comparison Between Moment Features and Using a Free Classification Service in Change-Robust Texture Recognition

Texture is one of the important characteristics used to identify objects or target areas in images. One of the difficulties in texture recognition is to ensure robustness against movement, rotation, and scale changes. In the past, texture recognition has been based on features derived from wavelet transforms and classified by neural networks or other classifiers. However, it is difficult to classify textures of different scales and rotations using only these methods. We propose a texture recognition based on moment features. Wavelet coefficients corresponding to each signal band are used to compute moment features. For classification, each moment feature is fed to a neural network. The recognition results are compared with Google Teachable Machine, a free classification service, to verify the superiority of our proposed method. It is confirmed that our proposed method improves the recognition rate under certain conditions.