An Improved Feature Extraction and Classification Using Neural Method for Accident Detection

In this paper, we put forward a  method to detect the accident and report it to the close by “Emergency  Service Provider”. This emergency provider then arranges for the necessary help. This is done by taking a real time video from a surveillance camera and then applying the following steps. First an input video is transformed into frames and then a preprocessing step is applied inorder to enhance the image. Features of the image are then extracted and this values are stored which is then given as input for the next step. This feature classification is done using Principle Component Analysis method(PCA). The next step feature classification is to train the system and provide the frame which consists of the image of the vehicle which has taken part in the accident. This stacked and extracted features is given as an input to neural network which is a method used for feature classification and then mail to the client. In future, haze eviction is done in the detected frame using guided filter and finally this frame is mailed to the emergency provider.