Error detection and elimination method in bullet comparison and matching

Bullet mark comparison is one of the key techniques in the discovery of evidences in gun related cases. A bullet mark comparison method base on statistical method was presented in this paper. First the bullet marks data are gathered according to generally used equipment. Then set up frame for the bullet and analysis some kinds of error types existing in measurement. There are many coarse data that cannot be used in matching process, so a preprocessing on the raw data is need. Here an abrupt detection method is adopted which can estimate the bound of the next measurement value according to current minus sequence. If the neighbor´s value is not between this estimation interval, there may be faults happening at some probability. Finally, a statistics modeling is set up to construct a cost function , which can be used to evaluation choose of error until the cost function get minimum value. This method can eliminate error efficiently. In 12 bullets, there are 2 bullets failed and 10 bullets successfully formed matching pairs. Experiment showed that this comparison method can solve comparison problems of bullet mark in an efficiency way.