CATCH-ALL 

Executive Summary 

Traffic congestion in Metro Manila can be attributed to the high volume of vehicles, poor public transport services, missing and incomplete road network, and inefficient transport and traffic management institutions. This can be generally categorized to: volume-based and behavior-based traffic congestions. The objective of this study addresses a problem in behavior-based traffic congestion which is the widespread disregard for traffic rules, especially, in a block intersection with high volume of vehicular traffic. The lack of traffic personnel and logistical resources and inefficiency of traffic monitoring and traffic law enforcement prompted the researchers to develop an automated contactless apprehension system that used vision-based systems and machine learning. The system is comprised of a video capture system, which is a network of roadside cameras connected to a remote server location that analyzed video information. The video analysis system consists of a vehicle detection and tracking, license plate localization and recognition, and traffic violations detection algorithms. Vehicle detection and tracking have 89.19% accuracy with 88.24% precision. The study focused on two specific traffic violations which are number coding and swerving violation detection. Number coding has 94.67% with 86.67% precision. Swerving detection has 96.67% accuracy. A database consisting of traffic violations data, such as traffic violation, plate number, date and time are recorded for file keeping and evidence. A notification system is also provided to increase public awareness (motorists and pedestrians) of the implementation of the automated contactless apprehension system in the vicinity. An LED board shows the real-time video monitoring of the vicinity with the list of plate numbers of captured traffic violators. 

The Problem 

Traffic congestion problem in developing metropolitan cities can be generally classified into two major problems: volume-based and behavior-based traffic congestion. Volume-based congestion is due to overcrowding of vehicles plying along the road. Overwhelming of the road capacity due to the tremendous number of vehicles causes traffic flow slowdown. Metro Manila’s traffic congestion problems are usually caused by large volume of private vehicles, ineffective and uncoordinated traffic management across 17 LGUs, illegal parking along the roadside, too much commercial activities on both sides of the roads, and lack of road planning. On the other hand, behavior-based congestion is caused by lack of discipline in road agents, such as drivers, pedestrians, and even traffic enforcers. This caused severe traffic congestion, especially in block intersections in Metro Manila. Some road accidents can be attributed to this behavior-based traffic-related problems. 

Behavior-based traffic congestion problem can be attributed to poor traffic law enforcement and lack of logistical resources. A 24 hours active monitoring of public roads and intersections is not easily achieved because the task is exhaustive to human operators. Contact-based traffic apprehension increases traffic blockages due to manual ticketing and traffic law enforcers are also prone to corruption. In terms of resources, traffic data from CCTV which monitors road activities takes large amount storage capacity and impossible to hold for long periods of time. There is a need to develop an automated vision-based contactless traffic violations apprehension system that can aid in traffic monitoring and traffic law enforcement that is tailored for Filipino driving behaviors. 

The Solution 

The CATCH-ALL system is an expert computer-vision system used for vehicle identification and tracking of traffic violations in real-time, see figure 1. It has 3 main sub-systems: video capture, video analysis, and output sub-systems. 

Figure 1 – CATCH-ALL System Architecture 

Figure 2 shows the network infrastructure layout of the system. Smart and low-cost IP cameras (figure 3), which served as the video capture system, are installed in the Taft Ave.-Estrada St. intersection (figure 4). The camera network is connected via fiber optics cable to the remote server location, see figure 5, which processed the visual information (video analysis system). Finally, the processed information, which are traffic violations committed in the intersection, are shown in an output LED screen that serves as the output system. 

Figure 2 – Network Infrastructure Layout 

Figure 3 – NI smart camera before deployment 

Figure 4 – Camera Locations 

Figure 5 – Remote server 

Target Market (Family): Video Surveillance 

Target Sub-Market (Line): Video Analytics 

Contact 

elmer.dadios@dlsu.edu.ph 

alexis.fillone@dlsu.edu.ph 

edwin.sybingco@dlsu.edu.ph 

laurence.ganlim@dlsu.edu.ph 

robert.billones@dlsu.edu.ph 

john.anthony.jose@dlsu.edu.ph