CONVOLUTION NEURAL NETWORK (CNN) BASED OBJECT DETECTION IN OCEANOGRAPHIC IMAGERY

Abstract

Author(s): Bhawanpreet Kaur,Khushbu Cheetu

The oceanography plays the vital role in the monitoring of the large scale areas from the satellite imagery sources. The ship vessel, debris, planes, crashed planes and boats detection, recognition and localization becomes very important for the surveillance of the multiple objects in the wider area like oceans. The satellite imagery collected from the satellite sources makes very larger volumes of data, which makes it very difficult for live monitoring by humans. Hence the requirement of automated surveillance mechanisms becomes higher for the segmentation, localization, detection and recognition of the objects in the ocean images. The proposed model has been designed by using the neural networks with the color and texture based features for the detection and analysis of the objects in the oceanographic imagery. The results have proved the proposed model’s robustness with only 8 negative cases out of 86 total cases, which makes the accuracy of the detection up to 90.32% overall.