Guo, Bin (The University of Texas at Dallas), Hansen, John (University of Texas at Dallas)

Enhanced Vehicle Detection System with Advanced Deformation Feature Extraction Algorithm

Scheduled for presentation during the Poster Session "Detection, estimatation and prediction for intelligent transportation systems" (WeAT17), Wednesday, September 25, 2024, 10:30−12:30, Foyer

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

Autonomous vehicles are at the forefront of advancing transportation technologies. Vehicle detection, essential for autonomous driving, remains a critical area with ongoing challenges despite extensive research. In this study, we propose a vehicle detection system designed to identify vehicles in complex scenarios. This system is based on the latest object detection model and is enhanced with a deformable convolutional algorithm to improve vehicle detection accuracy. The proposed system has three main contributions: First, it allows the network to adapt to geometric variations in input images, enhancing the detection and recognition of objects with various shapes and sizes. Second, it learns offsets to dynamically adjust convolutional kernels, improving focus on critical features and enabling more effective feature extraction and vehicle recognition. Third, it effectively handles features at different scales, which is essential for processing objects of varying sizes in images. We evaluated the proposed system using a naturalistic driving scenario dataset. The experimental results demonstrate that our system outperforms other baseline models in vehicle detection tasks.