利用卷積神經網路對空拍影像中車輛進行偵測與分類

CNN-based Vehicle Detection and Classification for Aerial Image

王鈞奕(Jyun-Yi Wang)

指導教授 ： 蘇志文

中原大學/電機資訊學院/資訊工程研究所/碩士(2018年)

https://doi.org/10.6840/cycu201800664

摘要

本論文中，我們著重於使用CNN架構解決空拍影像中的車輛進行偵測與分類問題。在智慧型交通運輸系統中，對都會區號誌化交岔路口中車輛進行偵測與分類，一直是個重要的議題。我們使用無人飛行載具搭載數位相機設備錄製影像，並將影像中車輛分為4種，公車、貨卡車、轎車、和其他大型車。使用了深度學習方法Mask R-CNN對車輛進行偵測並分類，之後找出車輛遮罩以擬合其矩行範圍，並且和其它深度學習方法進行了實驗比較。實驗結果顯示，在測試空拍影像中有0.76以上的平均精度均值。

關鍵字

深度學習 ； 卷積神經網路 ； 車輛分類 ； 車輛偵測

並列摘要

In this paper, we focus on the problem of vehicle detection and classification in aerial images by using CNN architectures. The detection and classification of vehicles at the intersection in urban area has always been an important issue in intelligent transportation system. We use high resolution digital camera on unmanned aerial vehicles to record aerial images. We adopt Mask R-CNN to detect and classify the vehicles into four types including buses, trucks, cars, and other. The detected mask of each vehicle will be fitted by a rotatable rectangle shape as the final result. The experimental results show that the mean average precision is outstanding in the test aerial image.

並列關鍵字

Deep learning ； Convolutional neural network ； Vehicle classification ； Vehicle detection

參考文獻

[1] A. Ajay, V. Sowmya, K. P. Soman, “Vehicle detection in aerial imagery using eigen features,” International Conference on Communication and Signal Processing (ICCSP), pp.1620-1624, April. 2017.

[2] N. Dalal, B. Triggs, “Histograms of Oriented Gradients for Human Detection,” IEEE Computer Society Conf. on Computer Vision and Pattern Recognition (CVPR), vol. 1, pp. 886–893, Jun. 2005.

[3] M. ElMikaty, T. Stathaki, ” Car Detection in Aerial Images of Dense Urban Areas,” IEEE Transactions on Aerospace and Electronic Systems, Volume: 54, Issue: 1, pp. 51-63, July. 2017.

[4] Z. Zivkovic, “Improved Adaptive Gaussian Mixture Model for Background Subtraction,” IEEE International Conf. on Pattern Recognition (ICPR), vol. 2, pp. 28-31, Aug. 2004.

[5] R. Girshick, J. Donahue, T. Darrell, “Rich feature hierarchies for accurate object detection and semantic segmentation,” Computer Vision and Pattern Recognition (cs.CV), pp.580-587, June 2014.