Application of adaptive and neural network computational techniques to Traffic Volume and Classification Monitoring
We are developing a Traffic Volume and Classification Monitoring (TVCM) system based on adaptive and neural network computational techniques. The value of neutral networks in this application lies in their ability to learn from data and to form a mapping of arbitrary topology. The piezoelectric strip and magnetic loop sensors typically used for TVCM provide signals that are complicated and variable, and that correspond in indirect ways with the desired FWHA 13-class classification system. Further, the wide variety of vehicle configurations adds to the complexity of the classification task. Our goal is to provide a TVCM system featuring high accuracy, adaptability to wide sensor and envirorunental variations, and continuous fault detection. We have instrumented an experimental TVCM site, developed PC-based on-line data acquisition software, collected a large database of vehicles` signals together with accurate ground truth determination, and analyzed the data off-line with a neural net classification system that can distinguish between class 2 (automobiles) and class 3 (utility vehicles) with better than 90% accuracy. The neural network used, called the Connectionist Hyperprism Classification (CHC) network, features simple basis functions; rapid, linear training algorithms for basis function amplitudes and widths; and basis function elimination that enhances network speed and accuracy. Work is in progress to extend the system to other classes, to quantify the system`s adaptability, and to develop automatic fault detection techniques.
- Research Organization:
- Los Alamos National Lab., NM (United States)
- Sponsoring Organization:
- Department of Transportation, Washington, DC (United States)
- DOE Contract Number:
- W-7405-ENG-36
- OSTI ID:
- 10191773
- Report Number(s):
- LA-UR-93-2855; CONF-940128-1; ON: DE93040256
- Resource Relation:
- Conference: Transportation Research Board annual meeting,Washington, DC (United States),Jan 1994; Other Information: PBD: [1993]
- Country of Publication:
- United States
- Language:
- English
Similar Records
Traffic operations, traffic signal systems, and freeway operations 1995. Transportation research record
Vehicle Localization in 3D World Coordinates Using Single Camera at Traffic Intersection
Related Subjects
99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE
TRAFFIC CONTROL
NEURAL NETWORKS
COMPUTER CALCULATIONS
MAPPING
PERSONAL COMPUTERS
DATA ACQUISITION SYSTEMS
SIGNAL CONDITIONING
MONITORING
EFFICIENCY
320203
990200
LAND AND ROADWAY
MATHEMATICS AND COMPUTERS