Automatic licence plate detection using adaptive neuro fuzzy inference system and wavelet packets

摘要

This paper presents an automatic licence plate detection (ALPD) system to improve the road safety and to monitor the vehicle movement on the streets. The system comprises a surveillance camera mounted at possible speeding points or at illegal parking zones which captures images and processes them to identify the errant vehicle owners for initiating appropriate legal action. The proposed scheme is cost-efficient and substantially reduces the manpower required to monitor the streets. ALPD uses a combination of K-Means Clustering and adaptive neuro-fuzzy inference system (ANFIS) to detect the Licence Plate from a given image. Wavelet Packet analysis and ANFIS are used to identify the alphanumeric characters. The scheme was tested on a sample of 52 test images pertaining to different configurations of the licence plate in terms of its location or font and using diverse colours of the vehicles, and was determined to have an accuracy of 98.3% and an average computational time of 900 ms. The algorithm was found to be robust, time efficient and has high recognition rate compared to many other schemes currently in practice. The use of Wavelet Packet analysis also makes the proposed scheme immune to variations pertaining to the font of the characters used for the licence plates.