Irregular Target Object Detection Based on Faster R-CNN

2018 4th International Conference on Environmental Science and Material Application

生态环境科学;材料科学

Zhang, Bin^1 ; Zhang, Yubo^1 ; Pan, Qinghui^1

Zhengzhou University, Zhengzhou

450006, China^1

关键词: Data set;    Learning network;    Network methods;    Network training;    Target data set;    Target detection algorithm;    Target feature;    Target object;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/252/4/042111/pdf
DOI  :  10.1088/1755-1315/252/4/042111

来源: IOP

【 摘 要 】

Forthe shortcomings of traditional target detection algorithms can only extract specific target features for detection, propose the Faster R-CNN target detecti-on model of deep learning, combined with VGG16 and ResNet101 convolutional neur-al network methods, to detection of irregular target objects. Experiments established two types of irregular target data sets, walnut and jujube, use the network training and testing, verified the feasibility of deep learning network for detecting irregular target objects. The experimental results show that the Faster R-CNN target detection networ-kof training on the self-built data set, the final detection result reaches 95%, which proves the effectiveness of the network for detecting irregular target objects.