Deep learning architectures for domain adaptation in HOV/HOT lane enforcement

Abstract

High Occupancy Vehicle (HOV) and High Occupancy Tolling (HOT) lanes have been commonly practiced in several jurisdictions to reduce traffic congestion and promote car pooling. Camera-based methods have been recently proposed for a cost-efficient, safe and effective HOV/HOT lane enforcement with the prevalence of video cameras in transportation imaging applications. An important step in automated lane enforcement systems is classification of localized window/windshield images to distinguish passenger from no-passenger vehicles to identify violators. This can be performed using deep convolutional neural networks (CNNs), which are shown to significantly outperform hand-crafted features in several classification tasks. Training/fine-tuning CNNs require a set of passenger/nopassenger images manually labeled by an operator, which requires substantial time and effort that can result in excessive operational cost and overhead. In this paper, we study adaptability of three popular CNNs (i.e., AlexNet, VGG-M, and GoogLeNet) across different domains in classifying passenger/no-passenger images as part of an automated HOV/HOT lane enforcement system. Our experiments over 40, 000 side-view vehicle images show many interesting insights for domain adaptability of these deep learning architectures.

VPDS: An AI-Based Automated Vehicle Occupancy and Violation Detection System

“…Parts of the proposed framework were earlier reported in (Xu et al 2014), (Artan et al 2016) and (Wshah et al 2016). This paper claims sufficient novelty and improvements over these previous works.…”

“…Figure 4 depicts that YOLOv3 detects the windshields accurately despite the presence of other similar entities like sunroofs, which in the case of DPM have to be removed manually by fine-tuning preprocessing parameters for each camera. This was one of the main drawbacks of the solution proposed in (Wshah et al 2016). Even after manual parameter tuning, the detection accuracy of DPM based models is lesser than YOLOv3.…”

“…(Xu et al 2014) uses DPM for localization and Support Vector Machines (SVM) for classification while (Artan et al 2016) uses DPM for windshields extraction from images and feature-specific models for classification. (Wshah et al 2016) proposes a solution for detecting whether there is a passenger present in a vehicle or not, which is essentially a binary classification and thus can only be used for figuring HOV2+ violations. Our work not only improves detection, but also counts the exact number of people in a vehicle by incorporating multi-class classification in the rear images.…”

Vehicle Type Detection by Convolutional Neural Networks

“…To this end a special kind of Convolutional Neural Network (CNN) has been used, namely AlexNet [5]. It has been previously used in traffic monitoring tasks [1,16]. The standard procedure to employ CNNs (AlexNet in particular) to object recognition, provides the neural network with an input image where the object of interest occupies most of the image area.…”

“…For example, it is allowed that trees partially block the sight of a vehicle [1]. In current applications of CNNs to vehicle detection, either the camera footage is manually segmented to ensure that these conditions are fulfilled [1], or a special sensor arrangement is implemented so that the acquired images always satisfy them [16].…”