Research On 3D Reconstruction And Visualization Of Ultrasonic Logging Image Based On Deep Learning And Virtual Reality

As one of the important technologies in the field of computer vision,3D reconstruction technology is widely used in building measurement,industrial manufacturing,medical imaging and other fields.Compared with two-dimensional data,three-dimensional data contains richer spatial information,and can accurately describe the external contour and surface features of the target object in three-dimensional space.In the exploration and development of oil and gas resources,the location and identification of cracks and holes has always been a key issue in the study of underground oil reserves.Ultrasonic imaging logging technology transmits ultrasonic waves to the borehole wall or the inner wall of the casing through the logging instrument,and obtains the echo signal according to the reflection characteristics of the acoustic wave.After the logging instrument processes the collected data,a two-dimensional amplitude image is drawn for researchers to analyze.Although the amplitude image obtained by ultrasonic imaging logging has the advantages of full wellbore coverage and wide application range,it lacks the spatial information to describe the wellbore structure,and researchers cannot efficiently observe and analyze the spatial structure of the wellbore.At the same time,the working environment of downhole high temperature and high pressure and full of mud and other media affects the performance of logging instruments and data storage equipment,which ultimately leads to the low resolution of the output ultrasonic logging image.Therefore,in view of the above problems,this paper proposes a three-dimensional reconstruction method based on deep learning and point cloud.The three-dimensional wellbore model with high resolution and rich texture is reconstructed,and a threedimensional wellbore observation platform based on Unity3 D is designed.Users can observe the high-quality three-dimensional wellbore model in a virtual scene and analyze its crack shape,hole distribution and wellbore structure.The main research contents of this paper are as follows:Firstly,aiming at the problem of low resolution of the original ultrasonic logging image,this paper uses the deep learning network ESRGAN to perform super-resolution reconstruction to obtain high-resolution ultrasonic logging images.In the training of the network model,through the principle of mutual game between the generator and the discriminator,the generator is prompted to continuously learn the image features in the ultrasonic logging image,so as to generate a high-resolution logging image closer to the real image,while the discriminator discriminates the input real image and the false image generated by the generator,and constantly adjusts the network parameters.Finally,the trained network is used to reconstruct the original logging image to obtain a highresolution image with an image size four times that of the original logging image.The experimental results show that the reconstructed ultrasonic logging image has higher resolution,richer texture details and more data.Secondly,aiming at the problem of three-dimensional reconstruction of ultrasonic logging images,this paper proposes a three-dimensional reconstruction method based on point cloud.Firstly,the pixel information is extracted from the high-resolution logging image,and the three-dimensional point cloud data including three-dimensional space coordinates and RGB color values is reconstructed by combining the measurement radius in the logging data.Then select the appropriate storage format to record the point cloud data,use KD-Tree to divide and sort the disordered point cloud,and calculate the point cloud normal for model rendering and surface reconstruction.Finally,the rolling ball method is used to triangulate the three-dimensional point cloud,and a closed and complete three-dimensional wellbore model is obtained.The experimental results show that the three-dimensional wellbore model obtained after point cloud surface reconstruction has high quality and clear surface features.Finally,aiming at the problem of visualization of 3D wellbore model,this paper designs a 3D wellbore observation platform based on virtual reality technology.The platform allows researchers to enter the virtual observation scene through VR equipment or keyboard input,and observe and analyze the three-dimensional wellbore model in an all-round way.