Dynamic Event Camera Object Detection and Classification Using Enhanced YOLO Deep Learning Architectur

Researchers have developed sophisticated computer vision applications for embedded platforms using the recent advancements in vision systems, such as distributed smart cameras. Important object detection techniques perform worse than standard object detection techniques when dynamic video data volume rises. Moving items blurring, moving quickly, occluding the background, or dynamic background changes in the foreground of a video frame can all be problematic. Salience detection is inadequate as a result of these problems. This work builds a deep learning model to address the issue. This work focuses on the three basic object detection, recognition, and classification processes with the goal of categorizing things that move before similar properties. The proposed method outperforms the conventional YOLO and FRCNN algorithms in classifying distant scenes with moving objects, as demonstrated by experimental results on the feature extraction and object classification public dataset and our dataset, which show 98.94% accuracy for the proposed method. With respect to accuracy, F-measure, mean absolute error (MAE), and computational load, Enhanced YOLO tests out existing video salient object recognition techniques on the benchmark dataset. In terms of accelerating and lowering computing burden, experiment shows that the YOLO method works better than other top-notch state-of-the-art methods.