Bayesian Unsupervised Vehicle Counting

This paper defines a novel task, unsupervised vehicle-counting from images, and presents the first practical solution that gives an accuracy comparable to supervised alternatives.

The major application of the task is Web-camera-based city traffic monitoring systems, where vehicle-counting must be done from very low-quality images. In this case, existing object-detection classifiers are impractical because of low resolution, poor viewing angle, and frequent occlusions. Also, the cost of preparing many training images is often prohibitive, since the quality of the images can be too low even for manual vehicle-counting. This calls for an unsupervised and robust approach to vehicle counting.

We formalize the problem as a task for Bayesian density estimation, where the number of vehicles is related to the total area of pixels that may correspond to vehicles in an image. We use the infinite Gaussian mixture model with a specific definition on the mean value with the framework of nonparametric Bayes and variational Bayes (VB) for model selection and computational efficiency. Using real-world Web-camera images, we show that the accuracy of the proposed approach is good enough for our application and robust for image quality. To the best of our knowledge, this is the first practical method for counting objects from images without training data.

By: Takayuki Katasuki, Tetsuro Morimura, Tsuyoshi Idé

Published in: RT0951 in 2013

LIMITED DISTRIBUTION NOTICE:

This Research Report is available. This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. It has been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g., payment of royalties). I have read and understand this notice and am a member of the scientific community outside or inside of IBM seeking a single copy only.

RT0951.pdf

Questions about this service can be mailed to reports@us.ibm.com .