26 September 2013 Highway traffic segmentation using super-resolution and Gaussian mixture model
Author Affiliations +
One benefit of employing computer vision techniques to extract individual vehicles from a highway traffic scene is the abundance of networked, traffic surveillance cameras that may be leveraged as the input video. However, the acquisition sensors that are monitoring the highway traffic will have very limited quality. Additionally, video streams are heavily compressed, causing noise and, in some cases, visible artifacts to be introduced into the video. Further challenges are presented by external environmental and weather conditions, such as rain, fog, and snow, that cause video blurring or noise. The resulting output of a segmentation algorithm yields poorer results, with many vehicles undetected or partially detected. Our goal is to extract individual vehicles from a highway traffic scenes using super-resolution and the utilization of Gaussian mixture model algorithm (GMM). We used a speeded-up enhanced stochastic Wiener filter for SR reconstruction and restoration. It can be used to remove artifacts and enhance the visual quality of the reconstructed images and can be implemented efficiently in the frequency domain. The filter derivation depends on the continuous-discrete-continuous (CDC) model that represents most of the degradations encountered during the image-gathering and image-display processes. Then, we use GMM followed by the clustering of individual vehicles. Individual vehicles are detected from the segmented scene through the use of a series of morphological operations, followed by two-dimensional connected component labeling. We evaluate our hybrid approach quantitatively in the segmentation of the extracted vehicles.
© (2013) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Amr Hussein Yousef, Amr Hussein Yousef, Jeff Flora, Jeff Flora, Khan Iftekharuddin, Khan Iftekharuddin, "Highway traffic segmentation using super-resolution and Gaussian mixture model", Proc. SPIE 8855, Optics and Photonics for Information Processing VII, 88550G (26 September 2013); doi: 10.1117/12.2026438; https://doi.org/10.1117/12.2026438

Back to Top