Irregular Shape Recognition Algorithm based on Rotation Invariant Features

In order to improve the shortcomings of the existing irregular shape recognition algorithms that has a large amount of calculation and poor resistance to scaling and rotation transformations, an irregular shape recognition algorithm based on rotation invariant features was proposed. A rotation-invariant direction vector was constructed through the central axis and mass points of the irregular shape. The projection of the irregular shape to the direction vector was calculated, and the rotation-invariant contour moment feature sequence was constructed with the average distance from the contour to the direction vector. Then the rotation invariant feature of the image was constructed by combining the projection sequence and the contour moment feature sequence. Finally, the dynamic time warping (DTW) algorithm was selected to measure the similarity of image feature sequences. The experimental results shows that the algorithm in the article has a small amount of calculation and time-consuming, meets the real-time requirements basically, has good robustness to image scaling and rotation transformation, and has certain resistance to slight perspective distortion of the image.