AUTOMATIC RECOGNITION ALGORITHM FOR BULLET ABRASION BASED ON FEATURE POINT EXTRACTION

摘要

The bullet abrasion recognition is important in the bullet tracking. A reasonable and efficient method is required to recognise and analyse whether the bullets shot from the same gun. In this paper, an algorithm for automatic detection of bullet abrasion is proposed. Firstly, ten bullet abrasion images were collected from HD microscope. Then the collected bullet abrasion data was de-noised by mean filter. The change of features points on the surfaces of each warhead section was extracted. A linear model was used to fit these features. The maximum value of the function gradient was identified. Finally, the similarities of two bullets abrasion were determined by correlation analysis to achieve a high accurate rate of recognition. It is validated through experiments that the new method is reliable, simple and efficient.