Improved Object Detection Algorithm using Ant Colony Optimization and Deep Belief Networks Based Image Segmentaion.

Abstract— Object detection is a very important application of image processing. It is of vital importance for object dynamic surveillance and other applications. So far, object detection has been widely researched. It shows an efficient coarse object locating method based on a saliency mechanism. The method could avoid an exhaustive search across the image and generate a small number of bounding boxes. After that, the trained DBN is used for feature extraction and classification on sub-images. This paper represents that the a variety of strategies based on object detection and efficiency of object detection framework using a saliency prior and DBNs for remote sensing images. This research works proposed an efficient object detection using the ant colony optimization and deep belief networks. The motivation behind the proposed approach is easy and efficient. Object Detection Algorithm .

Keywords—Obje4ct Detection, Deep Belief Networks, Object Detection Framework Ant Colony Optimization.

I. INTRODUCTION

After the unsupervised pretraining, a supervised layer is added to the top of the DBN to build a classifier. The probability distribution of the layer is defined as follows: ( ) ( ) ∑ ( ) which is also known as softmax regression [15]. P (class = j) is the probability that the data are assigned to class ( ) is the function of the model. At the fine-tuning stage, the back propagation algorithm is used to fine-tune the whole network until convergence. After training the deep model, detection is conducted for test images. 1.1 Deep Belief Networks A DBN is a multilayer generative model with several layers of restricted Boltzmann machines (RBMs), where by every level encodes precise dependencies one of several systems while in the level beneath it. Multilevel features could be extracted when each layer with the generative layer wise unsupervised learning algorithm. The model has been applied with success in a variety of computer vision tasks. DBN for aircraft detection in remote sensing images are utilized as training samples. 1.2 RESTRICTED BOLTZMANN MACHINE (RBM) Restricted Boltzmann machine (RBM), helpful to signify a single level of the model. Restricted Boltzmann machines are generally intriguing for the reason that inference is not difficult within them furthermore, as to remain effectively utilized while blocks with regard to exercising greater models. All of us initial prove that incorporating secret items brings stringently improved upon modelling electric power, when a second theorem shows that RBMs are generally common approximates connected with discrete distributions.
Read More