An Efficient Recognition Method for Object Detection System

In this paper, we proposed an efficient method of recognition for object detection systems. The proposed method uses convolution layer and PCA to quickly reduce dimensions and extract reliable features. In addition, features extracted from the proposed method calculate vectors in areas corresponding to background of the image. Called eigenvalues and used to create a mask for background. It splits the image into objects and backgrounds. To determine the efficiency of the proposed method, evaluate the INRIA benchmark image data set from $640\times 480$ to $128\times 64$ to measure accuracy performance. As a result, the performance showed high accuracy and low parameter memory. In conclusion, the proposed method confirmed that efficient object detection is possible.