Deep Learning Based Thermal Object Recognition under Different Illumination Conditions

Object recognition has recently a tremendous development because of the utilization of deep learning network. The recognition accuracy frequently experiences some sources of variation that can be found in images. Probably the most difficult variations are actuated by changing illumination conditions. Numerous researchers are targeting improving performance of object recognition frameworks towards various illumination conditions. The better model understanding and contributions of deep network in the object recognition motivate to utilize the deep networks for object recognition in different illumination conditions. This paper presents an accurate and proficient approach for recognizing the objects under varying illumination conditions. Hence, this proposed technique presents deep network-based system for illumination invariant object recognition in thermal image dataset. The thermal image database is utilized for the training and validation of deep network. In this work, Faster R-CNN is utilized to adequately recognize objects in thermal images under various illumination conditions. The results exhibit that the proposed work can significantly upgrade the recognition performance, outperforming other state-of-the-art strategies. This work can be an incredible assistance in driver assistance system and surveillance frameworks.