基于消失点检测算法的束靶红外图像畸变校正

传统的霍夫变换、Cannylines直线检测算法、霍夫概率变换方法在图像上的直线检测效果不佳,存在检测线段不连续不正确的问题,因而,利用Sobel滤波对红外图像横轴和纵轴两个方向分别进行锐化,通过线段检测(LSD)算法实现线段特征检测,进而经线段聚类拟合获得图像中完整的直线,通过对直线交点计算获得消失点,最后依据透视关系计算得到校正图像。实验结果表明,该方法可以实现对中性束红外图像的自动有效校正。

Neutral beam infrared image distortion correction based on vanishing point detection

The key parameters of the neutral beam, such as beam uniformity and beam divergence angle, can be obtained by analyzing the infrared images generated by the beam bombarding the target surface. Due to the camera setup angle, the IR images show geometric distortion, which affects the accurate analysis of the beam parameters. Therefore, so the images should be corrected for the distortion. The traditional Hough transform, canny lines algorithm, and probabilistic Hough transform methods are not effective in detecting straight lines on this image, and there is a problem of detecting discontinuous and incorrect line segments. In this paper, Sobel filter is used to sharpen the infrared image in the horizontal and vertical directions respectively and line segment feature is detected by line segment detector (LSD) algorithm. Then complete straight lines are obtained in the image by clustering and fitting the line segments according to the geometric and angular relationships between them. Vanishing points is calculated by the intersection points of the lines. Finally, the corrected image is obtained based on the perspective relationship. Experiment results prove that this method can achieve automatic and effective correction of neutral beam infrared images and it lays a basis for obtaining key parameters of the beam.