Wear particle shape feature based on spike-angle image and circular subband decomposition

According to the idea of wear particle spike parameter, the concept of spike-angle parameter and spike-angle image, and the wear particle image shape feature extraction method based on spike-angle image were proposed. Firstly, by calculating the spike-angle parameter of all the wear particle contour points at different step size, the spike-angle image can be acquired. The row of the spike-angle image reflects the global feature of wear particle contour, and the column of spike-angle image reflects the local feature of wear particle contour. Then, circular subband decomposition is applied to the rows and columns of spike-angle image separately, and the spectrum entropy of each subband is calculated. At last, each column of spectrum entropy matrix is quantized to normalized histogram, the global histogram feature and local histogram feature are used as the wear particle shape feature. The proposed wear particle shape feature extraction method makes full use of the global information, local information and multiscale information of wear particle shape, and has translation invariance, scale invariance, and rotation invariance. The experiment shows that the method proposed can be used effectively for wear particle recognition.