| Summary: | 碩士 === 中華科技大學 === 電子工程研究所碩士班 === 99 === Paul Viola and Michael Jones proposed a real-time object detection algorithm. The Paul Viola and Michael Jones algorithm using the Haar function and Adaboost algorithm for image identification method is by far one of the highest positioning accuracy algorithms. The calculation speed of this algorithm is much faster than other object detection algorithm. Although, the detection process of this algorithm is very quickly, but the training process require a lot of computation time to obtain image features to be used in detection process.
This research is to improve the Viola Adaboost algorithm to reduce the computation time needed for the training process of the algorithm with the similar accuracy. The series expansion of the template by taking the average the car license plates with Haar functions is first performed in this research. Then, the items of the series expansion with larger coefficient values are selected and deleted those items with small coefficient values. The training process of our proposed algorithm uses only the selected weak classifiers corresponding to the selected Haar functions, and therefore, the computation load of the training process can reduced. In this thesis, the experimental results show that the proposed algorithm can reduce the computation time and work well.
|
|---|
