Petroleum Casing Damage Intelligent Identification Based On Instance Learning

The monitoring of casing damage degree and type is always the key of the oilexploitation and well workover.The ultrasonic well logging image is widely used in casingdetection because the image is direct-viewing.The image needs to deal with the noise andimage segmentation,in order to be further applied to the subsequent processing of welllogging interpretation.The application of image processing and Machine learning to welllogging interpretation software improves the efficiency and accuracy of the interpretation andis of great significance.This article focuses on the problem of intelligent identification of casing damagetype,using biorthogonal wavelet transform multi-layer improved threshold denoising methodto denoise the logging image,using the improved geometric active contour model to segmentimage and extracting the characteristic of segmentation target, using ID3algorithm togenerate a decision tree,using the decision tree to identify casing damage type.The research work of this paper mainly includes the following three aspects:Firstly,this paper analyzes the wavelet transform denoising method,according to thedefects of standard orthogonal wavelet denoising and the calculation methods of hard and softthreshold,applying biorthogonal wavelet to image denoising,and improves the method ofthreshold calculation.The experiment shows that the denoising effect of this method isobvious, after denoising the signal-to-noise ratio improves.Secondly, this paper researches the geometric active contour model combined with thecurve evolution and level set theory,on the basis of C_V model,improved the energyfunction.The experiment shows that the segmentation effect of this method is better than thetraditional edge operator segmentation algorithm and C_V model.Thirdly, it analyzes lots of different types of casing damage images, extractingcharacteristic parameters which can be used to classify, using the ID3algorithm to trainsamples subset and generate a decision tree.The experiment proves that the classificationaccuracy of this tree can satisfy the practical requirements.Finally,Combined with theseresearch, which is applied in the casing type intelligent recognition system, and it has a goodeffect in practical application.