Detection transformer algorithm with efficient feature extraction for surface contaminant detection in a microsystem device

Aiming to address the detection problem of contaminants with diverse shapes in microsystem devices, this paper proposes a detection transformer algorithm with efficient feature extraction for surface contaminant detection in microsystem devices (EFE-DETR). This algorithm is based on the real-time detection transformer (RT-DETR) framework, and an efficient extraction module is constructed as the backbone network, enhancing the extraction ability and computational efficiency of key features of contaminants. The entanglement transformer block (ETB) is introduced at the end of the backbone network, the semantic information of contaminants is enriched, and a more comprehensive feature representation is formed. A recalibration feature fusion module is designed. Through multi-scale feature recalibration, the problem of information loss of contaminants’ features during transmission was alleviated. The experimental results show that compared with the RT-DETR model, the EFE-DETR model has significantly improved the three core evaluation indicators of mean average precision (mAP), precision (P), and recall (R), which provides a feasible solution for high precision and high efficiency automatic detection of surface contaminants in microsystem devices.