Airplane Detection and Classification Based on Mask R-CNN and YOLO with Feature Engineering

Attarian，Adam1; Luo，Minxuan2; Luo，Yangyang3; Tran，Hien4; Xu，Xingfei5; Yi，Chunxiao6; Yu，Yahe7; Zhang，Fa3

10.1007/978-3-031-16078-3_52

2023

2367-3370

2367-3389

Lecture Notes in Networks and Systems

543 LNNS

752-768

Deep learning algorithms achieve good performance in object detection and image classification. In this paper, we apply two algorithms, Mask R-CNN and YOLOv3, to the Rareplane dataset for airplane detection and classification. To achieve better performance in the fine grain classification problem, we propose a multi-step algorithm: Mask R-CNN is used to obtain bounding box, an edge extraction algorithm is used to get a more precise mask, the obtained masks are standardized, and their features are extracted. Using this algorithm, the mask type in the mask library with the most similar features is identified as the type of aircraft. Preliminary test results demonstrate that this algorithm is effective in fine grain classification, with an overall precision rate of 89.6% for the Airbus A300 and 88.6% for the Airbus A319.

Airplane classification Feature engineering Mask R-CNN YOLOv3

Others

English

2-s2.0-85138271240

Scopus

1.Pacific Northwest National Laboratory,Richland,United States
2.Southern University of Science and Technology,Shenzhen,China
3.Zhejiang University,Hangzhou,China
4.North Carolina State University,Raleigh,United States
5.Northeastern University,Boston,United States
6.China Agricultural University,Beijing,China
7.Dalian University of Technology,Dalian,China

Attarian，Adam,Luo，Minxuan,Luo，Yangyang,et al. Airplane Detection and Classification Based on Mask R-CNN and YOLO with Feature Engineering[C],2023:752-768.