Automatic Detection of Accidents under CCTV Monitoring

Abstract: Object detection followed by tracking structure which incorporates deep learning concept recognized as the Faster Regional Convolution Neural Network(Faster R-CNN) which is introduced and applied in order to detect accidents automatically when CCTV is being watched which is done in following three ways (1) Wrong-Direction (2) Stop (3) Person out of vehicle. This mechanism accepts a video input file as input which results the identification of different objects and each object identified with unique number. Thus this video is split into number of images and each image is recognized as a frame .These frames are then compared with previous frames using R-CNN algorithm to find out the accident image. This accident detection mechanism can find out accidents with minimal time. The proposed algorithm improves the results based on the quality of the video input.

Keywords: Faster R-CNN; CCTV; Object Tracking; Object detection; Deep learning; Image; Video