基于粒子群参数优化的多个自由平面标定法的图像畸变校正研究

发布时间：2018-04-19 03:12

本文选题：图像畸变校正 + 多个自由平面标定法　；参考：《昆明理工大学》2017年硕士论文

【摘要】：摄像机透镜晶体材料折射率等固有属性、透镜表面制作工艺导致光滑程度参差不齐以及人工组装等随机因素,不可避免造成成像时的几何畸变现象,影响了图像信息的原始性,当该图像用于后续处理时,不能得到对象的真实信息,因此研究畸变校正技术具有现实意义。SF6断路器是电力系统中的关键设备之一,电力规程需要定时人工现场巡检其气体压力、密度值,而压力表盘位置过高或位于夹层都会使得工作人员难以直观其指针示值,通过机器识别技术和遥视技术相结合,实现电力设备工作参数实时监测也具有重要意义。论文针对因摄像机镜头成像造成的图像几何畸变问题,研究了基于粒子群算法与多个自由平面标定法融合的图像畸变校正方法,并将其应用于工业图像机器识别中。分析了摄像机镜头畸变成因、畸变类型以及非线性畸变校正方法;采用多个自由平面标定法对摄像机数学模型参数进行标定;针对粒子群算法,给出了基于目标值的惯性权重因子计算方法;利用改进的粒子群算法优化了摄像机数学模型的标定参数,从而对非线性畸变图像进行校正。最后,分别用迭代法和Otsu算法对校正后的SF6断路器压力表盘指针图像进行阈值分割;并对比Sobel算子、Canny算子和形态学三种方法,对分割结果提取边缘信息;利用基于Hough变换的直线与圆提取方法定位指针,从而将指针指示角度变换为压力示值,实现了视感检测在电力系统智能遥视中的应用。实验结果表明,给出的基于改进粒子群算法参数优化的多个自由平面标定法的图像畸变校正技术能够较好复原畸变图像,预处理的结果为后续实现分割和边缘信息提取提供了保证。将视感检测融入电力系统中的"四遥"技术,丰富了遥视技术的内涵,替代了人工定时巡检,降低了远程视频监控人员的目测工作强度。
[Abstract]:The intrinsic properties such as refractive index of camera lens crystal material, uneven smoothness of lens surface and random factors such as artificial assembly, etc., inevitably lead to geometric distortion in imaging, which affects the originality of image information.When the image is used for subsequent processing, the real information of the object can not be obtained. Therefore, it is of practical significance to study the distortion correction technology. SF6 circuit breaker is one of the key equipments in power system.The electric power regulations require regular manual on-site inspection of the gas pressure and density values, and the high position of the pressure gauge disc or its location in the interlayer will make it difficult for the staff to visualize the pointer value, which can be combined with machine recognition technology and remote viewing technology.It is also of great significance to realize real-time monitoring of working parameters of power equipment.In order to solve the problem of image geometric distortion caused by camera lens imaging, the image distortion correction method based on particle swarm optimization (PSO) and multi-free plane calibration method is studied and applied to industrial image machine recognition.The causes of camera lens distortion, the type of distortion and the correction method of nonlinear distortion are analyzed. The mathematical model parameters of camera are calibrated by using multiple free plane calibration methods, and the particle swarm optimization algorithm is used to calibrate the mathematical model parameters.The calculation method of inertial weight factor based on target value is given, and the calibration parameters of the mathematical model of camera are optimized by using improved particle swarm optimization algorithm to correct the nonlinear distorted image.Finally, we use iterative method and Otsu algorithm to segment the corrected SF6 circuit breaker pressure gauge pointer image, and compare the three methods of Sobel operator Canny operator and morphology to extract edge information from the segmentation results.The method of extracting straight line and circle based on Hough transform is used to locate the pointer, and then the pointer indication angle is transformed into the pressure indication value, and the application of visual sense detection in intelligent remote view of power system is realized.The experimental results show that the proposed image distortion correction technique based on the improved particle swarm optimization (PSO) algorithm with multiple free plane calibration methods can recover the distorted image well.The preprocessing results provide a guarantee for the subsequent implementation of segmentation and edge information extraction.The integration of visual sense detection into the "four remote" technology in power system enriches the connotation of remote viewing technology, replaces the manual timing inspection, and reduces the visual intensity of remote video surveillance personnel.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP18

【参考文献】