A multi-target corner pooling-based neural network for vehicle detection
- PDF / 1,779,095 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 14 Downloads / 163 Views
(0123456789().,-volV)(0123456789(). ,- volV)
EXTREME LEARNING MACHINE AND DEEP LEARNING NETWORKS
A multi-target corner pooling-based neural network for vehicle detection Li-Ying Hao1
•
Jie Li1 • Ge Guo1
Received: 31 January 2019 / Accepted: 29 August 2019 Ó Springer-Verlag London Ltd., part of Springer Nature 2019
Abstract Convolutional neural network has shown strong capability to improve performance in vehicle detection, which is one of the main research topics of intelligent transportation system. Aiming to detect the blocked vehicles efficiently in actual traffic scenes, we propose a novel convolutional neural network based on multi-target corner pooling layers. The hourglass network, which could extract local and global information of the vehicles in the images simultaneously, is chosen as the backbone network to provide vehicles’ features. Instead of using the max pooling layer, the proposed multi-target corner pooling (MTCP) layer is used to generate the vehicles’ corners. And in order to complete the blocked corners that cannot be generated by MTCP, a novel matching corners method is adopted in the network. Therefore, the proposed network can detect blocked vehicles accurately. Experiments demonstrate that the proposed network achieves an AP of 43.5% on MS COCO dataset and a precision of 93.6% on traffic videos, which outperforms the several existing detectors. Keywords Convolutional neural network Vehicle detection Multi-target corner pooling Intelligent transportation system
1 Introduction Vehicle detection is the process of using sensors or computer vision technology to detect vehicles on the road. It is an important research topic of intelligent transportation system (ITS), which can offer reliable information for traffic management and control [1]. The urban traffic burden has increased dramatically, and it is urgent to establish an efficient ITS for traffic information collection, monitoring and management. In a traditional ITS, vehicle detection is always solved by employing special sensors. Jo and Jung [2] model and analyze optimal power-saving methodologies for an ultrasonic sensor and present a computationally efficient vehicle & Ge Guo [email protected] Li-Ying Hao [email protected] Jie Li [email protected] 1
College of Marine Electrical Engineering, Dalian Maritime University, Dalian 116026, People’s Republic of China
detection algorithm using ultrasonic data. And Kim et al. [3] propose a vehicle speed estimation using wireless sensor networks. These methods are unaffected by light or weather and can acquire the speed while detecting the vehicles. However, the traffic flow is usually interrupted by the installation of sensors, and the maintenance cost of these sensors is high. As the development of digital image processing, a vision-based vehicle detection system appears due to its great performance in detecting vehicle types, density, velocity and even predicting traffic accidents [4]. Unzuetal et al. [5] present an adaptive multicue background subtraction for vehicle det
Data Loading...
