A Context-Aware, Computer-Vision-Based Approach for the Detection of Taxi Street-Hailing Scenes from Video Streams

Area of Science:

• Computer Vision
• Human-Autonomous Taxi Interactions (HATIs)

Background:

• Autonomous taxis are increasingly deployed globally.
• Intuitive Human-Autonomous Taxi Interactions (HATIs) are crucial for adoption.
• Automated street-hailing recognition for autonomous taxis is underdeveloped.

Purpose of the Study:

• To propose a novel computer vision-based method for detecting autonomous taxi street hailing.
• To address the gap in automated street-hailing recognition technology.
• To develop a system that integrates visual and contextual information for enhanced detection.

Main Methods:

• Inspired by interviews with 50 experienced taxi drivers in Tunis.
• Distinguished between explicit and implicit street-hailing cases.
• Developed a computer vision pipeline using visual cues (gesture, position, head orientation) and contextual data (space, time, weather).

Main Results:

• The proposed method achieved satisfactory results in realistic settings.
• Accuracy: 80%
• Precision: 84%
• Recall: 84%

Conclusions:

• The integrated approach of visual and contextual information effectively detects taxi street hailing.
• The method shows promise for improving the user experience of autonomous taxis.
• Further development can enhance the reliability and efficiency of HATIs.