Research Of Compressed Sensing For Paper Defection Image Data Based On C++

With the rapid growth of demand for paper products in people’s daily life, productions of paper and paperboard also increase, leading to the paper machine more faster and the paper width more wider in paper-making industry. In order to ensure the quality of products, it is need to detect the surface defection in paper-making industry. The typically used method is called machine vision technology which acquires the paper image data in high speed motion with industrial cameras, and get the desired results with digital image processing. While as the substantial increasing in paper production, the amount of data to be collected and transmitted is also increasing significantly, which puts much pressure on the real-time processing system, and makes it easy to omit the detection of the paper that may have defection, also the stability of the system will be influenced. Therefore, the transmission problem of large data for paper defection detection system is increasingly serious and need to be solved as soon as possible.Compressed sensing is a new theory that guiding the acquisition and transmission for signals. It includes three aspects:the sparse representation for signals, the designation of measurement matrix and the reconstruction of signals. Compressed sensing can achieve collection and compression of signal at the same time, this will avoid the consumption of resource happened in the traditional mode which firstly collects large amounts of data and then drops much of redundant data. Using this feature, this thesis studied the transmission and reconstruction of the paper defection image on the framework of compressed sensing with the expectation that the data of transmission would be reduced effectively, and the real-time problem of the system would be solved. Based on this idea, following works was made:(1)A research on structures and working modes of existing paper defection detection systems was made.With the increasement of collected data,a transmission problem that all existing systems faced was pointed out.Compressed sensing was put forward to solve this problem. Specifically, one kind of software and hardware designation about paper defection detection syetem was studied and described. One achievement of compressed sensing was based on these foundations.(2)A theoretical research and simulation analysis about data processing of paper defection image was gaven using compressed sensing. Firstly, this thesis studied the compressibility of the paper defection image and decompose it on discrete cosine field and wavelet field, then the prerequisite for sparsity of paper defection image was verified. Secondly, the thesis gave some generating methods for some measurement matrixs, and the effection of reconstructed image on different measurement matrixs was studied. Lastly, this thesis analyzed the reconstruction problem of paper defection image and gave implementation steps for some commonly used algorithms. Also some simulation analysis for reconstruction effection on greedy algorithms and one kind of convex relaxation algorithm were made, the metrics were based on PSNR and running time and some reconstruction images were gaven as well.(3)Some C++achievements for application of compressed sensing on paper defection image were gaven, including the wavelet transform process of image, the generation of random measurement matrix, the algorithm of OMP. And also the thesis combined these C++functions with the whole software processing system.It is easy to find out that the transmission data is reduced effectively on the framework of compressed sensing, and the reconstruction images both have high PSNR and high subjective recognition rate.It is able to satisfy the requirements of subsequent processing.These results not only reduce the amount of transmission data but also ensure the accuracy of the reconstruction,at the same time,it has some adaptability for the tendency of paper machine with high speed and wide width and finally ensure the real-time detection of the paper image.