No abstract

Remainder particles in closed electronic equipment equipped in spacecraft are a potential threat to spacecraft. The volume of remainder particles is tiny, and remainder particles are similar to the general components of equipment in shape and size. So the current methods used to detect remainder particles cause false detection and missing detection seriously. In order to resolve above problems, a coarse- to- fine framework is proposed to detect remainder particles in closed electronic equipment of spacecraft, which includes a backbone network, a coarse classifier, and a fine detector. At first, we use ResNet-18 as the backbone network to learn the robust feature of remainder particles. Then, we design a coarse classifier. The feature maps containing remainder particles are filtered from it. Finally, the extracted patches containing remainder particles are sent to the fine detector for further location of remainder particles. The experimental results indicate that our framework exhibits superior performance on various evaluation indicators, especially reaching 92.54% on AP. It realizes efficient and accurate detection of remainder particles in closed electronic equipment of spacecraft.