Vehicle Detection in Aerial Images Based on Region Convolutional Neural Networks and Hard Negative Example Mining.

Tianyu Tang,Lin Lei,Huanxin Zou,Shilin Zhou,Zhipeng Deng

doi:10.3390/s17020336

Abstract

Detecting vehicles in aerial imagery plays an important role in a wide range of applications. The current vehicle detection methods are mostly based on sliding-window search and handcrafted or shallow-learning-based features, having limited description capability and heavy computational costs. Recently, due to the powerful feature representations, region convolutional neural networks (CNN) based detection methods have achieved state-of-the-art performance in computer vision, especially Faster R-CNN. However, directly using it for vehicle detection in aerial images has many limitations: (1) region proposal network (RPN) in Faster R-CNN has poor performance for accurately locating small-sized vehicles, due to the relatively coarse feature maps; and (2) the classifier after RPN cannot distinguish vehicles and complex backgrounds well. In this study, an improved detection method based on Faster R-CNN is proposed in order to accomplish the two challenges mentioned above. Firstly, to improve the recall, we employ a hyper region proposal network (HRPN) to extract vehicle-like targets with a combination of hierarchical feature maps. Then, we replace the classifier after RPN by a cascade of boosted classifiers to verify the candidate regions, aiming at reducing false detection by negative example mining. We evaluate our method on the Munich vehicle dataset and the collected vehicle dataset, with improvements in accuracy and robustness compared to existing methods.

Highlights

In very high resolution (VHR) remote sensing images, vehicle detection is an indispensable technology in both civilian and military surveillance, e.g., traffic management, urban planning, etc.vehicle detection from aerial images has attracted significant attention worldwide [1,2,3,4,5].automatic vehicle detection in aerial images still has a lot of challenges due to the relatively small size and variable orientation of vehicles (Figure 1a)
We adopt four widely used measures to quantitatively evaluate the performance of our method in the following—namely, recall rate, precision-recall curve (PRC), mean average precision, and F1-score
The mean average precision (mAP) metric is measured by the area under the PRC

Summary

Introduction

In very high resolution (VHR) remote sensing images, vehicle detection is an indispensable technology in both civilian and military surveillance, e.g., traffic management, urban planning, etc.vehicle detection from aerial images has attracted significant attention worldwide [1,2,3,4,5].automatic vehicle detection in aerial images still has a lot of challenges due to the relatively small size and variable orientation of vehicles (Figure 1a). In very high resolution (VHR) remote sensing images, vehicle detection is an indispensable technology in both civilian and military surveillance, e.g., traffic management, urban planning, etc. Vehicle detection from aerial images has attracted significant attention worldwide [1,2,3,4,5]. Automatic vehicle detection in aerial images still has a lot of challenges due to the relatively small size and variable orientation of vehicles (Figure 1a). Real-time detection in such large-scale aerial images with intricate backgrounds (Figure 1b) increases the difficulties. The existing vehicle detection methods in aerial images are mostly based on sliding window search and manual features or shallow-learning-based features [6,7,8,9,10,11]. The work of [2]

Methods

Results

Conclusion