| Magnetic Flux Leakage (MFL) inspection technology is the main method of on-line inspecting long distance oil-gas transportation pipeline. Because of the increasing of the inspection resolution, data precision and detection distance, existing devices can not satisfy the demand of an enormous amount of MFL data in the capability of data storage speed and capacity. Researching on the algorithms suited for MFL data compression and designing high speed MFL data acquisition and compression storage systems have become the key of developing the new generation of MFL inspection devices. Manual identification method can not satisfied the demand of a huge amount of MFL inspection data because of the slow speed and low precision. There are very urgent needs of research on the pipeline defects quantitatively intelligent identification technologies. Focusing on above-mentioned problems, the MFL Pipeline inspection data compression technologies and defects identification technologies are researched in this paper. The main work of the paper is as follows:(1) Taking a typical pipeline MFL inspection image as an example, the differences in futures and statistical characteristics between the pipeline MFL images and general digital images are analyzed. The statistical characteristics differences of the pipeline MFL images with different resolution and precision also are analyzed.(2) The application of prediction encoding methods in MFL image lossless compression is studied and a prediction model suitable for 8bits MFL images is proposed. The application of the integer wavelet transform and the embedded wavelet encoding algorithm in lossless compression of MFL image is studied, and got an improved listless SPIHT zero-tree encoding algorithms. Two sets of data lossless compression schemes for MFL images with different resolution and precision are designed. For 8 bits low resolution MFL images, a compression scheme combining the prediction encoding algorithm with the arithmetic encoding algorithm is used, and for 12 bits high resolution MFL images, they are firstly processed by prediction encoding and integer wavelet transforming methods and finally encoded with improved listless SPIHT zero-tree encoding algorithm.(3) A segmentation scheme for the defects in MFL images is proposed. A MFL image is segmented with the OTSU method after preprocessing, such as getting rid of lift-off, and then an complete defects image can be separated from the whole MFL image by dilation operation. Some characteristic parameters of defects such as the length and the width can then be picked up from the image.(4) The lossless compression method of MFL image based on region of interest (ROI) is studied. A minimal rectangle which includes the segmented defect region of ROI is used and only the data in ROI region is compressed and stored.(5) A finite element analysis software ANSYS is used to establish the model of the pipeline defects inspection devices, and the relation between MFL signals and the defect parameters is studied. BP neural network is used to recognize defects parameters for simulated defects and artificial defects. Recognition results basically fulfil inspection demand and are improved by fusing the axial MFL signal and radial MFL signal.(6) A scheme of MFL inspection data acquisition and compression store system base on a FPGA for the shortages of the existing pipeline MFL inspection devices is proposed, and the key parts such as multi-channels sampling control, prediction encode, integer wavelet transform, improved listless SPIHT zero-tree encode and file storage control of hard disk are studied.
|
PDF Full Text Request