Design And Embedded Implementation Of Automatic Detection Algorithm For Underwater Gas Leakage

With the development oil and gas resource in submarine applications,people pay more attention to the pipeline transportation safety.Once the oil and gas leakage occurs in the submarine transportation pipeline,it will cause direct economic and marine ecology losses.Therefore,the research on underwater gas leakage detection is of great significance.The existing imaging-based sonar system detection is mainly based on manual screening,in order to reduce labor costs and achieve automatic detection of leaking gas,In this paper,an embedded module with the function of automatic gas leakage detection is designed and completed.In this paper,by analyzing the imaging characteristics of multi-beam bathymetric sonar images,a set of automatic detection algorithm for underwater gas leakage is proposed.First of all,the median filter is used to suppress the noise in the image.In order to reduce the amount of calculation and remove the interference in the bottom area,the method of amplitude maximum and sliding window is used to realize the bottom detection.Sonar images often have the characteristics of low contrast,which is not conducive to image segmentation,so top-hat processing is used for image enhancement to improve the contrast.Using maximum entropy segmentation,the image is divided into foreground information and background information,and connected regions are marked to extract suspected regions.The image is restored to a fan-shaped image by the nearest neighbor interpolation algorithm,and finally the upward trend of the straight line is extracted by Hough transform to determine whether it is a gas leakage region.Through the test of 120 images,it has high accuracy.After simulation and verification in MATLAB environment,the algorithm is transplanted to the embedded system with C language.In the process of deployment,some algorithms are optimized to improve the operation efficiency.The median filter is implemented by a median selection method based on fast sorting.Bottom detection and top cap processing extract the maximum value from the template based on the reference value.In order to improve the efficiency of the maximum entropy segmentation algorithm,the formula is further deduced to reduce the logarithmic operation.In order to facilitate the implementation of C language,a connected region tag based on BFS algorithm is designed.Compared with MATLAB,the feasibility of self-defined function is verified.This paper also completes the embedded module based on VxWorks of which the display and control software based on MFC with communication protocol between PC.And TCP/IP.The embedded module mainly completes three functions: receiving image data,processing image data,and returning detection results.The scheduling mode of multi-task under VxWorks adopts the combination of priority preemption and exclusive resources,and the communication between tasks uses shared memory and binary semaphores.The display and control software simulates the multi-beam sounding sonar data.The result data is received by the main control.Through the joint test of the two modules,the functional integrity of the two modules is verified.