Object Detection Based on Multi Perception Fusion

Konferenz: ICMLCA 2021 - 2nd International Conference on Machine Learning and Computer Application
17.12.2021 - 19.12.2021 in Shenyang, China

Tagungsband: ICMLCA 2021

Seiten: 6Sprache: EnglischTyp: PDF

Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt

Autoren:
Zhou, Xiaolin; Cheng, Lei; Wen, Feng (College of Information Science and Engineering, Shenyang Ligong University, Shenyang, China)

Inhalt:
Based on the existing object detection methods, the recognition accuracy of small and medium objects is low, and the feature information of small and medium objects in the detection layer is not abundant, this paper proposes a object detection method based on multi perception fusion. The method design is based on SSD, introduces multi object detection fusion module, fuses the sensing information corresponding to different scale receptive fields, and fully extracts multiscale sensing information; then the bidirectional network module is introduced to fuse the high-level semantic information and the low-level detail feature information to enrich the feature perception information of the detection layer. The experimental results on MS COCO dataset show that the method can effectively improve the overall detection effect, especially for small and medium objects.