Information Enhancement And Location Of Direct Part Marking With Low Quality On Metal Surface

Data Matrix barcode can be used to store many kinds of information of the part, such as material, process, quality, circulation, usage and so on. Laser direct part marking technology doesn’t contact with workpieces in the whole processing. And because of the adjustable spot size, this technology has higher machining flexibility. Also it is environmental. It is widely used thanks to these advantages. Marking Data Matrix barcode on metal surface with laser direct marking technology can allow the tracking of the part through its full life cycle, which is of great significance for both industry and our life. Collecting images is the first step to read barcodes. But because of the illumination, carriers themselves, some human factors, the collected barcode images often have low quality, which makes the read of barcodes difficult. To solve these problems, this paper did some researches on the information enhancement and location of Data Matrix barcodes with low quality. Researches were carried out as following:In the first part, a summary of low quality images of Data Matrix barcodes was did. Then the low quality images were divided into two categories according to their factors. One is caused by illumination environment; the other is caused by the carriers themselves.In the second part, the information enhancement of barcode images with low gray value in local area was researched. Because part of the barcode is in shadow or doesn’t get enough illumination, the collected barcode image often has low gray value in local area, which has bad impacts on the readability of barcodes. This paper choos-es two approaches to solve this problem. The first approach is based on the histogram equalization, while the second based on Retinex theory. Poor detail enhancement and noise over-enhancing phenomenon are two most terrible shortcomings of traditional histogram equalization. Then a detailed evolution process of it was discussed, and two most excellent algorithms, POSHE and CLAHE, were selected among those six algo-rithms. The algorithm based on iterative calculation and the algorithm based on center around had the best effect among all Retinex algorithms. This paper discussed the working principles of these two algorithms in detail. An important conclusion was obtained through subjective and objective evaluation. In terms of contrast enhance-ment, the order of these algorithms from high to low was POSHE, McCann, CLAHE, SSR. In terms of PSNR, the order of these algorithms from high to low was CLAHE, SSR, McCann, POSHE.In the third part, highlight area reconstruction of Data Matrix barcodes marked on metal surface was researched, and a five-step-reconstruction model based on local and global features was proposed. The reconstruction gist was introduced, which is the basis of the model. First, slant correction was carried out by orientating the "L finder pattern" right on the lower-left corner of the image. Then, each Data Matrix barcode module was located by mesh generation. Third, the highlight area was found by the Modified Specular-Free (MSF) image. Finally, the lost information in the high-light area was reconstructed by the proposed five-step-reconstruction model. The ver-ification experiment shows that this algorithm can remove the highlight area reliably and make more Data Matrix symbols to be read more easily.In the last part, location of Laser-directly-marked Data Matrix symbol with compli-cated background was researched, a location algorithm based on improved Edge Ori-entation Histogram (EOH) feature and line detection was proposed. The algorithm can be reliably used to locate Data Matrix symbols with different kinds of complicated backgrounds accurately, such as images containing other characters or patterns, mate-rial texture, abrasion or oil contamination or rust and so on. Firstly, the symbol image was divided into several subblocks with the same size. Secondly, improved EOH fea-ture and Support Vector Machine (SVM) were used to classify these subblocks. Thirdly, the barcode subblock area was centered to determine the line detection area. Lastly, Hough transformation was introduced to detect the finder pattern of Data Ma-trix symbol. Compared with other existing texture analysis methods or geometrical location methods, the proposed algorithm in this paper has advantage in both accuracy and average elapsed time.