Efficient Bayes Saliency-Based Object Detection on Images Using Deep Belief Networks

Abstract

Object detection has been one of the hottest issues in the field of remote sensing image analysis. The purpose of object detection is to find precise locations of the objects, one location at a time or all locations of all objects in an image. However, most current object detection methods developed earlier demonstrate unsatisfactory results. Therefore, this paper presents efficient Bayes saliency-based object detection on images using deep belief networks. First, a new Bayes saliency detection approach is presented in which prior estimation, feature extraction, weight estimation, and Bayes rule are used to compute saliency maps. In particular, an efficient coarse object locating method is used based on a saliency mechanism. Then, an efficient object detection framework is implemented which combines the unsupervised feature learning of Deep Belief Networks (DBNs) and visual saliency. After that, the trained DBN is used for feature extraction and classification on sub images. The method could avoid an exhaustive search across the image and generate a small number of bounding boxes, which
can locate the object quickly and precisely. Comparative experiments are conducted on the dataset and result analysis demonstrate that the accuracy and efficiency of our method than state of-the-art methods in terms of various evaluation metrics. Furthermore, this object proposal approach can improve the detection performance and the speed of several detection approaches.