| As the widely application of PDC (Polycrystalline Diamond Compact) bit, it becomes difficult to identify the lithology of cuttings with traditional ways. While cutting lithology is a crucial parameter for geology description, especially in logging process, hence it is necessary to seek a new method to identify the cutting lithology. The digital image processing technology has been used in many areas. While in petroleum industrial and cutting logging field, its function has been rarely explored. In this thesis, several digital image processing methods are used to extract the features of cuttings. Based on multi-characteristics extracting, an optimal procedure for cuttings'lithology identification has been determined, on the purpose of providing technical support for the ongoing project of cuttings description system developed in our laboratory.According to the method of the mode identification, the thesis carries out the research of identifying program for the lithology of cuttings through the process of obtaining image, extracting characteristics, classifying and identifying. In the process of research, the system for the experiment of shooting images of cuttings is built through the researcher's designing, so that the images of cuttings are obtained. All sorts of parameters of the system for shooting are confirmed via modifying the resolving power and the luminosity. All the images obtained have been processed and analyzed by applying two software systems (the software system for experimenting and the one of logging equipment of cuttings for testifying in the locale) both built by the team, as well as by compositing the program of the kernel module of the two systems.In the software system for experimenting, the thesis applies Fourier Transform, Wavelet Transform and Gabor Transform three methods of extracting characteristics and then carries out the groping for identifying with the Bayes statistical probability classifier. According to the analyzing of the experiment, each identifying accuracy is obtained. Appropriate characteristics are selected by referring to the box pictures of the statistics. Combining the selected characteristics produced by Fourier Transform and Wavelet Transform, the identifying accuracy of the mud stones and the sand stones is increased up to 80%. Thereinto a method about the Algorithm of Circular Fourier Spectral Energy(ACFSE) used for extracting characteristics of cuttings is designed based on the theory of Fourier Transform.In the software system of logging equipment of cuttings for testifying in the locale, using early experiences for reference, the thesis adopts the Sum and Difference Histograms, Fourier Transform and Gabor Transform three methods of extracting characteristics, as well as the Artificial Neural Network (ANN) classifier. For obtaining optimal identifying results, through all sorts of combination of the three methods and the modulating of the parameters of the classifier in the same time, a lot of experiments have been finished. The outcome indicates that using the Artificial Neural Network classifier, in the circumstance of that the number of the nerve sell of the covert layer is the number of characteristics plus 1, the identifying results of the lithology of cuttings is optimal employing the integrated method of combining the characteristics of both the Sum and Difference Histograms and Gabor Transform. Depending on this scheme, the identifying accuracies of mud stones and sand stones are attained as 100% and 89.2%, which can meet the need of testifying in the locale. So this identifying method of combining the characteristics of both the Sum and Difference Histograms and Gabor Transform, and making use of the Artificial Neural Network classifier, can be the optimal selection for the scheme of the description of the lithology of cuttings in the thesis.Aiming at the need of practicability, the thesis also puts forward the direction of further progress in the aspects of the system of obtaining images and the method of experiment. |
PDF Full Text Request