Research On 3D Scene Reconstruction Technology Based On Sequential Bladder Endoscopic Images

Optical bladder endoscopy is the gold standard for bladder cancer diagnosis.It plays an important role in disease diagnosis,surgical guidance and cancer monitoring.But because the imaging range of the endoscope probe is too small,the image magnification and field of view incompatibility,and inability to correlate individual 2D images with 3D structure form in the current field of vision,which limits its utility for quantitative or longitudinal studies of disease physiology or cancer surveillance.To solve the above problems,this paper proposes a three-dimensional scene reconstruction method based on sequential bladder endoscopy images.The main contents include the following parts:1.To solve the problem that traditional calibration methods can not be applied in human bladder,an endoscope self-calibration algorithm based on Kruppa equation is adopted.According to the projection property of absolute quadratic curve,a virtual calibration block is established to determine the endoscope's internal parameters and distortion coefficient,and to correct the image distortion.2.For endoscope image acquisition to the image quality is not high,uneven illumination,noise and other issues,a proprietary color adjustment algorithm processes is proposed.Firstly,the region of interest(ROI)of the corrected bladder image is selected,the ROI image is transformed from RGB color space to LAB color space,and the color enhancement is carried out by the adaptive histogram equalization algorithm with limited contrast,the bilinear interpolation algorithm is used to accelerate the process.Finally,by compared the number of feature corners of bladder image before and after pretreatment,the effectiveness of the pretreatment process is verified.3.The incremental Structure from motion algorithm was used to restore the three-dimensional point cloud of bladder.Firstly,the preprocessed bladder image is extractedand matched by SIFT algorithm,and an adaptive termination threshold RANSAC error matching rejection algorithm is proposed.By continuously correcting the interior point ratio p in the sampling process,the sampling times are determined adaptively,which avoids the problem that the sampling times and termination conditions of the traditional RANSAC algorithm are difficult to determine.Then the three-dimensional point clouds and camera posture of the initial image are restored by polar geometric constraints and triangle method,and camera pairs are added in incremental order,and the parameters are optimized by beam adjustment method to recover the three-dimensional point clouds of the bladder and the motion trajectory of the endoscope.The experimental results show that the average re-projection error of reconstructed bladder model data and standard clinical data is less than one pixel,which proves that the proposed algorithm is feasible.