License Plate Detection using Gabor filter banks and texture analysis

Abstract

This work discusses the detection and isolation of license plates, especially focusing on photos taken with a standard digital photo camera. The paper will describe a specific process for the detection of license plate using a range of morphological operations, the Gabor-filter and Gabor-filter bank, as well as preprocessing with the image segmentation methods. The features of the license plate, the height and width of the plate and the height and width of the characters will serve as the basis for forming structural elements with the morphological operations. The method of analysis of connected components is used for the determination of the exact position of the plate.

Comparison of Gabor Filter Bank and Fuzzified Gabor Filter for License Plate Detection

“…These cameras are considerably more expensive than commercial automatic digital cameras. REG cameras combine special optics with software support that contributes significantly to the development of high-quality initial images [1], [5].…”

“…When a Gabor filter is applied to an image, it gives the highest response at points and at edges where texture changes. By virtue of these characteristics, algorithms based on Gabor filters are successfully applied in computer vision applications [31] such as license plate detection, face detection, texture extraction, character recognition and others [1], [5].…”

“…After preliminary consideration, a function (x,y,f,Φ,σ) which represents the Gabor filter with the orientation of 0° where the f is the spatial frequency and the Φ denotes the filter orientation is obtained as follows [1], [5]:…”

“…There are numerous algorithms with an excellent performance that can be easily adapted to any license plate recognition system [2]. However, the extraction of a license plate from an image still remains a highly researched issue within object recognition, with many papers still being published on the topic [1], [3], [4], [5].…”

“…Wellconceived pre-processing step can compensate for flaws in an image acquisition, or ambient conditions, facilitating easier extraction of a license plate from an image in the later stages. The first algorithm used a Gabor filter bank with distinctly crisp parameters (denoted GFBank) [5], while the second algorithm relied on a fuzzified Gabor filter bank (denoted FuzzyGF) [1] with the parameters fuzzified to maximally adjust the filter's response. The problem of filtration using Gabor filters has been accessed by taking into account fuzzy logic, a powerful mathematical apparatus [6], [7], [8] that closely imitates human reasoning.…”