ISHD: Intelligent Standing Human Detection of Video Surveillance for the Smart Examination Environment

Abstract

In the environment of smart examination rooms, it is important to quickly and accurately detect abnormal behavior (human standing) for the construction of a smart campus. Based on deep learning, we propose an intelligent standing human detection (ISHD) method based on an improved single shot multibox detector to detect the target of standing human posture in the scene frame of exam room video surveillance at a specific examination stage. ISHD combines the MobileNet network in a single shot multibox detector network, improves the posture feature extractor of a standing person, merges prior knowledge, and introduces transfer learning in the training strategy, which greatly reduces the computation amount, improves the detection accuracy, and reduces the training difficulty. The experiment proves that the model proposed in this paper has a better detection ability for the small and medium-sized standing human body posture in video test scenes on the EMV-2 dataset.