Simultaneous Object Detection and Association in Connected Vehicle Platform

The connectivity in vehicular network extends the sensing capability in both the range and quality of the sensing data. One of the most significant benefits is the availability of sensing data from connected vehicles. Leveraging this data in useful ways is an attractive research topic in the communi...

Full description

Saved in:
Bibliographic Details
Published in2018 IEEE Intelligent Vehicles Symposium (IV) pp. 840 - 845
Main Authors Guo, Rui, Keshavamurthy, Shalini, Oguchi, Kentaro
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.06.2018
Subjects
Bandwidth
Detectors
Feature extraction
Image color analysis
Object detection
Task analysis
Online AccessGet full text

Cover

More Information
Summary:The connectivity in vehicular network extends the sensing capability in both the range and quality of the sensing data. One of the most significant benefits is the availability of sensing data from connected vehicles. Leveraging this data in useful ways is an attractive research topic in the community. In this paper, a novel one-pass deep neural network is proposed to implement object detection and the association simultaneously. Considering the bandwidth limitation in the typical vehicular network communication, the proposed algorithm not only highly compresses the feature representation but also maintains the high quality in the detection and association performance. The learning architecture is delicately designed to enhance the task by incorporating multi-modality features. Each modular unit in the system can be appropriately deployed in the vehicle onboard electrical control unit (ECU) and on remote servers to realize a pratical implementation in many applications.
DOI:10.1109/IVS.2018.8500420