Moving Object Detection and Localization in Camera-Trap Images

Simsek, Emrah; Ozyer, Baris; Özyer, Gülşah Tümüklü

doi:10.18185/erzifbed.509571

Cited by 1 publication

(2 citation statements)

References 25 publications

(12 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is well known that YOLO is a convolutional neural network that consists of convolution, pooling, and fullyconnected layers [29]. In a traditional convolutional neural network, , , is the feature value of the m layer n feature map of the position (i, j) calculated as following [30],…”

Section: B Object Detection Based On Yolomentioning

confidence: 99%

“…where , , is the feature map, and refers to the local neighborhoods of the position (i, j). In the last stage of the network, several fully-linked layers are used to convert 2D feature maps into a 1D feature vector [29]. In this output layer, the SVM algorithm which can be generally combined with the softmax operator or CNN is used for classification.…”

Section: B Object Detection Based On Yolomentioning

confidence: 99%

See 1 more Smart Citation

Vision-Based Moving UAV Tracking by Another UAV on Low-Cost Hardware and a New Ground Control Station

2020

View full text Add to dashboard Cite

Automatic flying target detection and tracking in video sequences acquired from a camera mounted on another Unmanned Aerial Vehicle (UAV) is a challenging task due to the presence of nonstationary cameras in the system, dynamic motion of the moving target, and high-cost computation for realtime applications. In this paper, our aim is to automatically detect and track moving UAV by another one while simultaneously flying in the air. In order to provide efficiently in real-time applications, we develop a vision-based low-cost hardware system integrated with an independent ground control station. We initially created a new public dataset called ATAUAV that includes different types of UAV images obtained from videos recording in our experiments and searches on Google Images for the training process. Deep learningbased YOLOv3-Tiny (You Only Look Once) is used for target detection with the highest accuracy and fastest results. Kernelized Correlation Filter (KCF) adapted with YOLO, which runs on low-cost hardware, is used for real-time detected target tracking. We compared the performance of the proposed approach with different tracking algorithms. Experimental results show that the proposed approach provides the highest accuracy rate as 82.7% and a mean fps speed as 29.6 on CPU. The dataset can be downloaded at http://cogvi.atauni.edu.tr/ResearchLab/PageDetail/Our-ATAUAVs-Dataset-86. INDEX TERMS Artificial neural networks, computer vision, kcf, object detection, object recognition, target tracking, unmanned aerial vehicles, yolo This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.

show abstract

Section: B Object Detection Based On Yolomentioning

confidence: 99%

Section: B Object Detection Based On Yolomentioning

confidence: 99%