Abstract
Object detection, as a fundamental task in computer vision, has been developed enormously, but is still challenging work, especially for Unmanned Aerial Vehicle (UAV) perspective due to small scale of the target. In this study, the authors develop a special detection method for small objects in UAV perspective. Based on YOLOv3, the Resblock in darknet is first optimized by concatenating two ResNet units that have the same width and height. Then, the entire darknet structure is improved by increasing convolution operation at an early layer to enrich spatial information. Both these two optimizations can enlarge the receptive filed. Furthermore, UAV-viewed dataset is collected to UAV perspective or small object detection. An optimized training method is also proposed based on collected UAV-viewed dataset. The experimental results on public dataset and our collected UAV-viewed dataset show distinct performance improvement on small object detection with keeping the same level performance on normal dataset, which means our proposed method adapts to different kinds of conditions.
Highlights
Object detection in Unmanned Aerial Vehicle (UAV), as a kind of burgeoning technique, has numerous applications, such as aerial image analysis, intelligent surveillance, and routing inspection [1,2,3,4]
Accuracy and processing time trade-off is important to real-world applications. Encouraged by these problems, we develop an object detection method based on You only look once (YOLO) [16], which focus on small object detection, named as UAV-YOLO
Following the UAV-viewed data classification method introduced in Section 3.3, YOLOv3 is trained as described in Section 4.1 and tested on UAV-viewed test dataset
Summary
Object detection in Unmanned Aerial Vehicle (UAV), as a kind of burgeoning technique, has numerous applications, such as aerial image analysis, intelligent surveillance, and routing inspection [1,2,3,4]. With the development of large-scale visual datasets and increased computation power, the deep neural network (DNN)—the convolutional neural network (CNN) [5]—has demonstrated record breaking performance in computer vision tasks including object detection [6,7,8]. It is still a challenging work due to special perspective. With the development of computation hardware, deep-learning-based methods can enhance the accuracy and realize real-time detection
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
