Bar Code Detection In Complex Background

With the advent of the age of Intemet of Things, as the key technologies of identifiaction in which., bar code technology will also get an unprecedented development. Especially two-dimension code has a wide range of applications in transport and logistics, identification, advertising, marketing and e-commerce and many other aspects, it is the most economic, practical automatic identification technology. Combining with the mobile phone will make the bar code market more prosperous in the future. But due to the ever-changing bar code format and complicated background, it is difficult to detect the barcode.Currently there is not a good method to accurately position the bar code, which has limited the future application of the barcode.This paper has a deep research in the bar code detection method,and summrises the commom problems in the general algorithms:(1)could not detect variety of barcode or can not detect multiple bar code simultaneously;(2)can not detecte the barcode in the complex background;(3)cannot effectively detect the barcode in media t besides printing on paper. This paper presents a new method for detecting barcode based on Adaboost,which effectively solves the problems above.The main content of this paper is as follows:(1)dividing the input image into sub-blocks to build samples,which makes the feature extraction and classification more simple and convenient.(2)analysis image texture feature.the paper extracts the image feature using the local Binary Patterns (LBP)and Gabor filter for image feature, and reduce the feature dimension.(3)In training the algorithm divides the input training samples into sub-blocks, and extracts LBP and Gabor filter feature as a vector which can describe the texture characteristic for each sub-block. Then the algorithm takes the feature vector as input, using new weight update mode and weak classifier bulid methods Adaboost to build a strong classifier.(4)In detecting module, divides the input image into sub-blocks, and extracts a feature vector which can describe the texture characteristic for each sub-block. Then the algorithm takes the feature vector as input, classifies the sub-block with the strong classifier above,. Finally, after the post process, we combine the barcode sub-block, get the localization result with rectangle.The experments shows that the algorithm can get a satisfactory result.