Study on Visual Perception-based Obstacle Detection in Platform Gap of Rail Transit
This study proposes a novel method for detecting obstacles in the gaps of rail transit platforms,which is based on the conventional residual convolutional neural network architecture and centered on a visual perception mechanism.By integrating visual perception strategies,this method significantly enhances the representational power of visual features and effectively overcomes the challenges posed by image variations due to changes in lighting and train vibrations.Experi-mental results from real-world scenario testing have confirmed the simplicity and effectiveness of this method,demonstra-ting its ability to significantly improve the detection accuracy of obstacles in platform gaps and potentially reduce safety risks to vehicles and passengers.
NETL
NSTL
万方数据
