Abstract
This paper develops a model to determine and to locate the optimal numbers of license plate recognition (LPR) to minimize error rates of O-D matrix estimation, percentages of LPR coverage (a proxy of installation cost), and percentages of recorded trips (a proxy of privacy invasion). A genetic algorithm is proposed to solve the combinatorial problem. To demonstrate the applicability of the proposed model and solving algorithm, two exemplified cases and a real-world case are investigated. The results have consistently showed that the optimal locations of LPR are at both ends and in the middle of the segments of a freeway corridor with heavy link traffics. If an extensive coverage of LPR is attempted, however, additional LPR may be placed at the segments with light link traffics so as to balance out the privacy invasion.
