Neural network based auto-scoring system for shooting range

• Main Author: Hou, Dajun.
• Other Authors: Song, Qing
• Format: Theses and Dissertations
• Published: 2008
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Computer hardware, software and systems; DRNTU::Engineering::Electrical and electronic engineering::Applications of electronics
• Online Access: http://hdl.handle.net/10356/4378
• Institution: Nanyang Technological University

In this thesis, a Computerized Auto-scoring System based on image processing and pattern recognition is presented. The scheme, which is implemented with the hardware system consisting of high-resolution digital cameras and personal computers, gains the advantages of low cost and easy maintenance that are two main requirements of an Auto-scoring System. Meanwhile, the system can achieve satisfactory accuracy and efficiency by using advanced pattern recognition technologies. Three kinds of classification methods — Statistical Classification, Radial Basis Function (RBF) neural networks and Support Vector Machines (SVM) — have been experimented for the particular problem called Bullet Hole Recognition in the system. All three methods have been tested based on the same samples and features. Experimental results show that both RBF and SVM can perform very well with error rate 1.85%. Thus, a function-well neural network based auto-scoring system for shooting range is built.