A New Method for Bar Code Localization and Recognition

A new method is proposed for the localization and recognition of a one-dimensional bar code. First, we eliminate most of the irrelevant image background and preserve the bar code area by using the Sobel edge detector, and then the image is divided into several horizontal strips of images in the same size. Because of the disadvantages of Hough transform of large computation-load and great time-consumption in detecting lines in an image, we propose a new and fast method for detecting lines and obtaining their inclination angles. Since bar codes consist of parallel black bars and white bars, and the edge points are continuous, we compute the angle of fitting lines by using the turning points, and considered this angle the inclination angle of the bar code. The area of lines with the same or similar angles is considered the area of bar code. If the angle is not 90 degrees with respect to the horizontal direction, rectify the bar code by bilinear interpolation. Recognizing the picked up bar code area is done by calling the decoding procedure which employs the similar-edge theory. MATLAB simulation results show that when applying the method to our image database, the localization rate reaches 99%, and the decoding rate reaches 98.5%.