Video Stream Mining for On-Road Traffic Density Analytics

Video Stream Mining for On-Road Traffic Density Analytics

Rudra Narayan Hota, Kishore Jonna, P. Radha Krishna
DOI: 10.4018/978-1-61350-056-9.ch011
OnDemand:
(Individual Chapters)
Available
$37.50
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Traffic congestion problem is rising day-by-day due to increasing number of small to heavy weight vehicles on the road, poorly designed infrastructure, and ineffective control systems. This chapter addresses the problem of estimating computer vision based traffic density using video stream mining. We present an efficient approach for traffic density estimation using texture analysis along with Support Vector Machine (SVM) classifier, and describe analyzing traffic density for on-road traffic congestion control with better flow management. This approach facilitates integrated environment for users to derive traffic status by mining the available video streams from multiple cameras. It also facilitates processing video frames received from video cameras installed in traffic posts and classifies the frames according to traffic content at any particular instance. Time series information available from various input streams is combined with traffic video classification results to discover traffic trends.
Chapter Preview
Top

Introduction

The speed and precision of natural vision system for living beings (human, animal, birds or insects) is amazing, yet less explored because of complexity involved in the biological phenomena. Every intelligent system (Intelligent robotics, Intelligent Traffic system, Interactive medical applications and human intension recognition in retail domain etc.) in many industries is attempting to simulate natural vision system. The major hurdles in such process are high computational complexity because of very high dimensional images data and the semantic gap between the image content and the observed concepts from natural images/scenes. Recent progress in computational power and understanding of local and global concepts in images opens path for new line of work in dynamic automation. Table 1 summarizes some of the emerging applications in various domains and existing challenges in vision based solutions.

Table 1.
Vision based application in different domain and issues
DomainApplicationsGeneral functions in computer vision systemsIssues of Computer vision Systems
1. Health CareComputer-aided diagnosis, surgical applications, Mammography Analysis, Detection of Carcinoma tissue, Retrieval of similar diagnosed images.i. Image acquisition: (Sensors- light, ultra sonic, tomography, radar)
ii. Pre-processing (Re-sampling, Noise reduction, Enhancement, Scale normalization)
iii. Feature extraction (Lines, edges, interest points, corners, blobs, color, shape and texture)
iv. Detection/Segmentation (region of interest, foreground and background separation, Interest points)
v. High-level processing (Object Detection, Recognition, Classification and Tracking)
i. Various types of image and videos (binary, gray, color), different data types (GIF, BMP, JPEG and PNG), and sizes (SQCIF, QCIF, CIF, 4CIF).
ii. Camera Sabotage (FOV obstruction, sudden pan, tilt, zoom) and Discontinuity in video streams
iii. Illumination (varied intensity and multiple source of lights)
iv. Blurring
v. Occlusion
vi. Different object size
vii. Changing Field of View in moving cameras
2.TransportSmall to large vehicle detection, Vehicle count, Traffic density estimation, Incident detection, Traffic rule violation detection, Eye and head tracking for automatic drowsiness detection, Lane/Road detection etc.
3. Security SurveillancePeople detection and tracking, Abnormal behavior recognition, Abandoned Objects, Biometric pattern recognition (Face, Finger prints), Activity monitoring in mines etc
4. ManufacturingCamber measurement, Item detection and classification, and Vision-guided robotics etc.
5. RetailCart detection, Vegetable recognition etc.

Complete Chapter List

Search this Book:
Reset