Date of Publication


Document Type

Master's Thesis

Degree Name

Master of Science in Manufacturing Engineering

Subject Categories



Gokongwei College of Engineering


Manufacturing Engineering and Management

Thesis Advisor

Robert Kerwin C. Billones

Defense Panel Chair

Elmer P. Dadios

Defense Panel Member

Alexis M. Fillone
Edwin Sybingco


Transportation is often viewed as crucial to a nation's economic growth as its efficiency is one of the metrics used to measure economic output and quality of life. However, traffic congestion disrupts mobility in major metropolitan areas, especially road-based systems. The government's solution to the need for improved traffic management is creating a local public transportation route plan (LPTRP), a highly technical procedure requiring mastery of one's accessibility and mobility requirements. Standard buses are at the top of the DOTr's hierarchy of transport service characteristics since they can deliver the passenger capacity per hour for the route in question. Therefore, it is essential that bus transportation planning and management reflect real-world data and traffic circumstances. The majority of mandatory surveys indicated in the LPTRP depend on two types of information: 1) the number of passengers boarding and alighting at each stop and 2) the location and speed of the bus throughout its journey. The surveys are conducted manually by observers who log data continually, posing a threat to data integrity due to the possibility of human mistakes. In addition to data integrity difficulties, the present techniques demand significant amounts of time and people, making them exceedingly inefficient.

This study aims to design and develop a smart vision-based bus passenger counter that can monitor the number of boarding and alighting passengers at specific times and places through edge computing and computer vision by utilizing the existing surveillance cameras onboard the bus. In addition, a web-based application is developed to visualize the data from the smart passenger counters. Tiny-YOLOv7 and FASTMOT were utilized in this study to detect and count the number of boarding and alighting passengers inside the bus. Results show that the trained Tiny-YOLOv7 achieved a 90.4% accuracy at 23.5 FPS, while the FASTMOT attained an accuracy of 93.33%. The developed web application was able to track the GPS location and display visualizations of the gathered data.

With the proposed system, traffic management teams of LGUs can easily monitor the status and movement of the public utility buses. Furthermore, data such as utilization ratio, passenger load, boarding, and alighting information, time of arrival and departure, and travel time can be derived from the gathered information. This can be used in the creation of the local public transport route plan as detailed in the guidelines provided by the Department of Transportation.

Abstract Format







Intelligent transportation systems; Computer vision; Bus occupants; Bus lines; Global Positioning System

Upload Full Text


Embargo Period