Accurate and fast infrared (IR) foreground object detection is one of the most significant issues to be solved due to its important meaning for IR target recognition, IR precise guidance, IR video surveillance, and so on. A common approach for such tasks is "background subtraction," which aims to detect foreground object through background modeling. Thus far, many background subtraction methods have been proposed and have achieved good performance. However, due to the special characteristics of IR images, a few algorithms are suitable for IR foreground object detection. Recently, features learned from convolutional neural networks (CNNs) have demonstrated great success in many vision tasks, such as classification and recognition. In this letter, we propose a novel multiscale fully convolutional network architecture for IR foreground object detection. Given a CNN model pretrained on a large-scale image data set, our method takes output features from different layers of the network. With features from multiple scales, our feature representation contains both category-level semantics and fine-grain details. The experimental results on IR image sequences show that the proposed method achieves the state-of-the-art performance while operating in real time.
Zeng, D. D.,Zhu, M.. Multiscale Fully Convolutional Network for Foreground Object Detection in Infrared Videos[J]. Ieee Geoscience and Remote Sensing Letters,2018,15(4):617-621.
APA
Zeng, D. D.,&Zhu, M..(2018).Multiscale Fully Convolutional Network for Foreground Object Detection in Infrared Videos.Ieee Geoscience and Remote Sensing Letters,15(4),617-621.
MLA
Zeng, D. D.,et al."Multiscale Fully Convolutional Network for Foreground Object Detection in Infrared Videos".Ieee Geoscience and Remote Sensing Letters 15.4(2018):617-621.
修改评论