![]() However, their application in realistic scenarios has not been ideal. CNNs have proved to be superior in terms of speed and accuracy when tested in challenging benchmarks. In the field of object detection and tracking, Convolutional Neural Networks (CNNs) have been widely used in computer vision applications ( Wu et al., 2017), and its effect is greatly superior to that of traditional detection methods, such as Fast R-CNN ( Girshick, 2015). Second, sensor-based localization technology is used to identify the lecturer’s position. First, a deep neural network is used to track the lecturer, and the camera is panned to capture the lecturer. This PT camera-based solution integrates two types of core technology. To solve these problems, we investigated the use of a PT camera enhanced with extra localization technology for the real-time tracking of the lecturer, even if it moved quickly beyond the camera’s field of view. However, in reality, the sudden movements of the lecturer often lead to tracking failures with such cameras. Some automatic recording systems are designed to capture the lecturer via an automatic PT camera ( Zawadzki & Gorgoń, 2015). Using a Pan-Tilt (PT) camera is more conducive to expanding the capture range. Using a fixed camera limits the lecturer’s range of movement and requires human intervention during the recording process. To produce high-quality lecture recordings using an automatic recording system, artificial intelligence technology is needed to direct the camera’s rotation because speakers move around a great deal during their presentations. Hence, automatically detecting the lecturer and recording video is necessary from a cost point of view. Moreover, the core of the online video recording is the camera operated by professional videographers ( Lalonde et al., 2010), which causes the large consumption of human-power. Most of the videos on online course platforms are recorded by technical experts in professional studios using expensive photographic equipment ( Tan, Kuo & Liu, 2019). ![]() Lecture recording plays an important role in online learning and distance education. Although the rapid development of computer technology has made it possible to provide richer educational resources online, the cost of designing and developing multimedia teaching facilities is still very high ( Zhang et al., 2021). Significantly, the advantages of the online teaching modes are gradually showing under the case of COVID-19 outbreak in recent years. Various online courses have emerged as information technology in education develops rapidly. ![]() The experimental results show that our system outperforms the systems without a PDR module in terms of the accuracy and robustness. ![]() We built the entire lecture recording system from scratch and performed the experiments in the real lectures. Further, when the lecturer is beyond the camera’s field of view, the PDR auxiliary module is enabled to capture the object automatically. In addition, to improve face detection performance on the edge side, we utilize the OpenVINO toolkit to optimize the inference speed of the Convolutional Neural Networks (CNNs) before deploying the model. First, the particle filter algorithm is used to fuse the PDR information with the rotation angle information of the Pan-Tilt camera, which can improve the accuracy of detection under the tracking process. This article proposes a lecturer tracking system based on MobileNet-SSD face detection and Pedestrian Dead Reckoning (PDR) technology to solve this problem. The key of the automatic recording system is lecturer tracking, and the existing automatic tracking methods tend to lose the target in the case of lecturer’s rapid movement. Automatic lecture recording is an appealing alternative approach to manually recording lectures in the process of online course making as it can to a large extent save labor cost. ![]()
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |