Intelligent Detection in Less Visibility by Saliency Techniques and Faster Region-based Convolutional Neural Networks
Decrease in visibility causes several difficulties. Many classic object detection techniques do not give satisfying results in less visibility. It is essential to detect and recognize the entities present in an image under these conditions, and to devise a better object detection mechanism. This paper solves this problem by proposing a solution which uses a multi step approach to obtain the desired results. The approach discussed in the paper usessaliency techniques and conventional object detection algorithms. The distorted image is enhanced via a deep neural network for visibility enhancement. The image frame with better quality undergoes saliency techniques so that less visible objects can be identified. An object detection algorithm, faster region-based convolutional neural network (R-CNN) is then run to yield bounding boxes for all the objects, including those having low visibility. The coordinates of the bounding boxes are then recorded and superimposed on the original distorted image having less visibility.