文献类型：期刊文献

英文题名：PointDet++: an object detection framework based on human local features with transformer encoder

作者：Tang, Yudi[1]; Wang, Bing[1]; He, Wangli[1]; Qian, Feng[1]

机构：[1] Key Laboratory of Smart Manufacturing in Energy Chemical Process, Ministry of Education, East China University of Science and Technology, No.130 Meilong Road, Shanghai, 200237, China

年份：2023

卷号：35

期号：14

起止页码：10097

外文期刊名：Neural Computing and Applications

收录：EI(收录号：20220811681630)

语种：英文

外文关键词：Chemical detection - Chemical plants - Feature extraction - Gesture recognition - Hydrogen peroxide - Object detection - Object recognition - Signal detection - Signal encoding

摘要：Object detection algorithm plays an important role in the field of chemical plant safety identification. In the field operation chemical plant scene, usually the targets to be detected are highly correlated with people, and most of them are small objects because of the long shooting distance. Conventional object detection algorithms use strong backbone module to obtain global features, which makes the algorithm module perform well in large targets. But these methods are difficult to be applied in small target detection scenes as they lack the use of local features. How to make better use of local features is the key to small target detection tasks. It is not only necessary to extract the local features from the original data, but also to consider the location relationship between them. To solve this problem, we propose a new object detection framework named PointDet++. The first step is to use the trained pose estimation model to obtain the local features of human body. Then, we reconstruct local features and global features, respectively, with transformer encoder and graph convolution. In the output layer, we integrate local features and global features according to the target to be detected, so as to improve the detection performance of our proposed model. Specifically, our framework significantly outperforms state of the art by 10.3 AP scores on field operation dataset in chemical plant. ? 2022, The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature.