Efficient SSD: A Real-Time Intrusion Object Detection Algorithm for Railway Surveillance

Abstract

With the rapid development of the Chinese Railway, the research of lightweight and real-time railway perimeter intrusion detection algorithm is crucial for railway security. A single shot multibox detector (SSD) can detect intruding objects but cannot be deployed to the non-Graphics Processing Unit (GPU) equipment of the railway system to achieve real-time detection. In this paper, EfficientSSD, a real-time lightweight object detector is proposed for railway surveillance detection. Based on the conventional SSD model, EfficientSSD replaces VGG16 with EfficientNet-B3 as the backbone network. The detector was trained on the Pascal VOC dataset and then on a Railway dataset with the mAP of 76.6%. EfficientSSD runs at about 11 FPS on a CPU <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">a</sup> with an overall accuracy of 96.95%. Compared with the SSD, our method obtains a precision improvement of 4.8% with only 1/2x model size and 2x speed.