Abstract:
Rainy days with ponding water on the road surface can cause vehicle crashes, injuries, and deaths. Surveillance video-based splashed water detection is a promising way to reduce traffic accidents. Surveillance videos have many lighting changes, illumination conditions, and complex backgrounds, making automatic recognition difficult. This paper proposes a deep learning-based splash detection method. We believe this is the first deep learning-based study on this topic. SWNet, a novel semantic segmentation network, extracts splashed water regions. An encoder-decoder structure captures water splashes. Pooling indices and using the light-weight decoder make SWNet efficient. SWNet’s multi-scale feature fusion structure integrates coarse semantic and detailed appearance information, improving accuracy and edge segmentation. To handle the unbalanced distribution of splashed water and backgrounds, a weighted cross entropy loss is used. A splashed water attention module integrates global contextual information in semantic segmentation to focus on salient regions of moving vehicles and splashed water. The proposed method outperforms state-of-the-art methods in experiments on a new splashed water dataset.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
