多尺度光斑中心的快速检测

发布时间：2018-11-10 12:21

【摘要】：分析了光斑图像成像特点和理想光斑灰度分布模型,针对含有多个不同尺度光斑的图像,提出了一种可以在复杂环境下一次性快速检测出多个光斑中心的方法。该方法基于高斯模糊后光斑中心不变的性质,先对含有大量光斑的图像进行快速多级高斯模糊,构建其高斯尺度空间;然后,使用加速的非极大值抑制方法在尺度空间内寻找多个尺度的局部极值,初步确定各光斑中心的像素级坐标;最后,联合这些坐标的邻域像素,拟合局部曲面,得到光斑中心的亚像素级精确位置。利用仿真实验和实物实验验证了提出方法的有效性。结果表明:该算法对640pixel×480pixel图像,处理时间仅需50ms,每千个光斑的平均检测时间为23ms,在复杂环境下正确率可达89%。此外,该方法对弱光斑较敏感,适合快速处理含有大量不同尺度光斑的图像,并能够有效减少光斑的错检和漏检。由于检测速度快,自适应性强,在实际应用中取得了良好的检测效果。
[Abstract]:In this paper, the imaging characteristics of spot image and the ideal gray distribution model of light spot are analyzed. A method to detect the center of light spot in complex environment is proposed for the image with different scales. This method is based on Gao Si's property that the center of light spot is invariant. Firstly, the image containing a large number of light spots is blur by a fast multilevel Gao Si blur, and then the size space of the image is constructed. Then, the accelerating non-maximum suppression method is used to search for the local extremum of multiple scales in the scale space, and the pixel level coordinates of each spot center are preliminarily determined. Finally, by combining the adjacent pixels of these coordinates, the local surface is fitted to obtain the accurate sub-pixel position of the spot center. The effectiveness of the proposed method is verified by simulation and physical experiments. The results show that the processing time of the algorithm is only 50 Ms for 640pixel 脳 480pixel images, and the average detection time for every 1000 spots is 23 Ms, and the correct rate can reach 89 Ms in complex environment. In addition, this method is sensitive to weak spot, and is suitable for fast processing of images with different scales, and can effectively reduce spot error and miss detection. Because of fast detection speed and strong adaptability, good detection effect has been obtained in practical application.
【作者单位】：大连理工大学海岸与近海工程国家重点实验室;大连理工大学电子信息与电气工程学部;
【基金】：国家自然科学基金资助项目(No.61202253)
【分类号】：TP391.41

【相似文献】