Automated class identification of modes of travel in shared spaces: a case study from India

Abstract

This paper presents a classification approach for road-user modes of travel. The classification does not assume well organized, and lane disciplined traffic. Instead, it relies on specific characteristics intrinsic for each road-user to predict the corresponding class. The classification relies on extracting the geometric and movement characteristics of road-users. As such, it is possible to classify road-users in shared space facilities and sites with high level of non-compliance. The classification is a multi-step procedure. First, movement features are used to discriminate between motorized and non-motorized road-users. Then, complementary features based on road-user geometry are added to differentiate between vehicles, rickshaws, powered two-wheelers, and buses. Experiments are performed on a video data set from a shared facility in New Delhi, India. A performance analysis demonstrated the robustness of the proposed classification method with a correct classification rate of up to 90 percent. By considering the movement attributes, the approach is tolerant to considerable variations in road-user physical details which often arises from choices of camera positions and partial occlusions. The research is part of the long-term goal to develop an automated video-based road safety and data collection system for developing countries.