Institute of Transportation, MOTC

• Date：2023-09-20
• Update：2023-09-20
• Department：IOT

Project Overview
      Advanced Driver Assistance Systems (ADAS) have become increasingly common in recent years. An increasing number of transportation companies have installed ADAS-related equipment in their vehicles to alert drivers to potentially hazardous situations and assist them in avoiding collisions. This 4-year plan and freeway bus operators have collaborated to collect massive driving images, driver behaviors, and driving safety warning records in order to optimize the application of ADAS system in safety analysis. In addition to analyzing image recognition technology to analyze vehicles, aberrant events inside vehicles, and find aberrant events that actually pose safety risks from a large number of warnings, the hot zone where aberrant events occur from a spatial standpoint was discussed. Plans for behavior-oriented high-risk road section improvement were developed and integrated into safety management systems (transportation operators).

      In the plan's first year (2021), image recognition technology for abnormal events was developed. Only 23% of aberrant events were discovered from 2,531 ADAS warnings using actual driving images of freeway buses outside vehicles and ADAS warnings as the basis. In the second year of the plan, images of vehicles inside vehicles were further integrated to develop aberrant event image recognition technology inside cars, observe drivers' hands, heads, and trunk dynamics, thereby analyzing potential distracting behaviors and the impact of various ADAS warnings and aberrant events while driving. Research results can help transportation operators gain insight into the behavioral characteristics of drivers and accidents, allowing them to take appropriate actions and solve problems.
Research Results
       4. Collaborate with highway bus operators to collect images inside and outside vehicles, driver behaviors, ADAS warning events, and other data. The Institute of Transportation has established a four-year plan for collecting natural driver data.
       5. To investigate potential distractive behaviors of drivers and the impact of potential distractive behaviors on aberra n t eve n t s , image recognition technology for aberrant events while driving in vehicles has been developed.
       6. Integrate images from inside and outside vehicles, driver behaviors, ADAS warning events, external road geometry, environmental
data, and potential distractive behaviors of drivers, and investigate the characteristics of frequent occurrences of aberrant events.

Result Promotion and Benefits

• Image recognition technology's phasal results have been applied to detect anomalous events both inside and outside of vehicles. The characteristics of frequent occurrences of aberrant events have been analyzed to serve as a reference for automobile transportation operators and a foundation for developing other safety improvement strategies.
• Participated in the digital transformation of the automobile transportation industry and a seminar for transportation safety scholars and experts; invited competent authorities and transportation operators to participate; and promoted the plan research results on January 18, 2023.

Summary of Research Results

The demonstration of image recognition technology development for abnormal events inside vehicles

Research Result Report
• Applying Artificial Intelligent Method for Exploring Risk-prone Road Section (1/4) -Driving Behavior Analysis and Image Recognition Technique Development for Aberrant Events (Published in June 2022)
• Applying Artificial Intelligent Method for Exploring Risk-prone Road Section (1/4) -Driving Behavior Analysis and Image Recognition Technique Development for Aberrant Events (Published in May 2023)