Visual Based Voice Control Navigation Technique for Visually Impaired Using AI Board

In this study we describe a strategy for many objects whose number is unknown and varies during tracking. The multiple objects tracking approach retains many assumptions about the object’s quantity and trajectories in the video based on preliminary results of object detection in each image, which may be missing and/or inaccurate detection. The best explanation for the video is chosen based on the image data, which also drives the process of extending and pruning the graph. The application is designed to aid visually impaired individuals by providing an auditory representation of detected objects. For this study recursive and non-recursive background modelling techniques were used for extracting foreground pixels from the each input video frame data. By using Non-recursive technique, background model data were estimated through sliding window approach. This approach mainly includes selective video frame storage in a separate buffer from the input video. In order to estimate the background model data, pixel extraction and size variation is done from the stored frame data. The main advantage of using Sliding window approach is for its accuracy in localization of pixel.