Intersection Traffic Count and Prediction

Lamia Iftekhar

 

Task

This project is an application of supervised learning algorithms. We propose to design a traffic counter that can estimate the traffic volume at an intersection based on data captured on a traffic camera. We also seek to build a model to predict the traffic volume at that intersection at any given time during a weekday.

Dataset

The dataset used for this project is obtained from the Hanover Police Department. It is a collection of still images taken from a single traffic camera mounted on the Lyme/ North Park intersection near Dartmouth Medical School. The database contains images from weekdays of two consecutive weeks during Jan/Feb 2012. They are sampled fifteen minutes apart from 8am to 5pm each day.

Figure 1: A sample image from the dataset

Method

Because the images are taken from a fixed point, the area covered in each image remain the same. We plan to pre-process the images so that our program can only focus on the relevant road area while learning to detect vehicles. Many projects dealing will fisheye images tend to flatten the images before working with them [1], a path we may take too depending on the complexity and the time-expense of the attempt. We will take a subset of these processed data as training sets, label the vehicles and use a boosting learning algorithm to perform vehicle detection. At this point, Adaboost seems like the best choice [2].

Once we have our traffic counter, we can use regression techniques to find a model that can predict the traffic volume at the intersection at any time of the day over a weekday.

Timeline

Dates below denote completion dates of their corresponding tasks.

20th of April Pre-processing
4th of May Implement boosting algorithm to train and test
8th of May Project milestone write-up and presentation
11th of May Perform regression and establish prediction model. Evaluate performance
31st of May Revise project, final write-up, prepare poster

References

  1. In Jung Lee; , "De-blurring algorithm for fish eye Image on CCTV with nonlinear equation," Control, Automation and Systems, 2008. ICCAS 2008. International Conference on , vol., no., pp.1986-1989, 14-17 Oct. 2008
  2. Gwang Yul Song; Ki Yong Lee; Joon Woong Lee; , "Vehicle detection by edge-based candidate generation and appearance-based classification," Intelligent Vehicles Symposium, 2008 IEEE , vol., no., pp.428-433, 4-6 June 2008