Deep Learning Based Motion Target Detection Algorithm
Xizhou Wang · 2024
With the dramatic growth of video data, the storage and computational resources required to process this huge amount of data have increased significantly. In order to cope with this challenge, it is necessary to extract the key information in the video in a more intelligent and efficient way, while filtering out a large amount of redundant content. In this paper, the traditional CNN model and Transformer model are constructed respectively using video frames of car motion process from video viewpoint as a dataset. The model performance is improved by advanced data preprocessing operations. The bilateral filtering technique is introduced in this study, aiming to improve the image quality and enhance the image processing effect through denoising operations, making it more applicable to the subsequent processing steps. Finally, the Transformer model is verified by the model and the recognition accuracy of the Transformer model is up to about 90%.