基于农作物图像序列的拼接算法研究

发布时间：2018-05-08 09:08

本文选题：图像配准 + 区域检测　；参考：《沈阳理工大学》2017年硕士论文

【摘要】：为了获得满足一定分辨率要求的大场景的图像,常常需要将关于场景的多幅序列图像进行拼接,图像拼接算法是实现这一目的的关键技术。该类算法具有很强的实用性,广泛地应用于遥感、医学等领域,尤其在农业应用领域具有广阔的应用前景,对于农业作物的监测、评估有着重要意义。本文主要是针对农作物序列图像的拼接算法进行深入的研究。本文首先介绍了图像拼接算法的研究背景和当前的研究现状,对各类算法的特点进行了比较分析,并介绍了在拼接过程中涉及到的一些关键技术。然后以基于特征匹配的图像拼接算法为研究方向,重点分析了SIFT和SURF特征提取的原理和特点,并在此基础上提出了结合感知哈希重合区域检测的拼接算法,设计了一种重合区域搜索算法,首先对待拼接的相邻序列图像进行重合区检测,确定有效的拼接区域,然后在有效区域内提取特征点及描述子实现配准。实验结果表明,该算法能够显著提高匹配速度和效率,提取稳定准确的特征点,减少误匹配,与现有算法相比有更好的实时性。在图像融合输出方面,针对配准图像可能出现的拼接缝明显、重影、亮度不均衡等问题进行了深入的分析,提出了基于重合区渐变系数融合和基于动态规划的最佳缝合线算法等解决方案,对拼接缝隙进行处理。对于序列图像存在的亮度差异问题,通过建立相邻序列图像的像素分布模型对图像的亮度进行均衡。实验结果表明,本文算法能明显改善融合图像的视觉效果,得到理想的图像输出。
[Abstract]:In order to obtain large scene images which meet certain resolution requirements, it is often necessary to mosaic multiple sequential images about the scene. Image mosaic algorithm is the key technology to achieve this goal. This kind of algorithm has strong practicability and is widely used in remote sensing, medicine and other fields, especially in the field of agricultural application. It is of great significance for the monitoring and evaluation of agricultural crops. This paper mainly focuses on the mosaic algorithm of crop sequence image. This paper first introduces the research background and current research status of image stitching algorithms, compares and analyzes the characteristics of various algorithms, and introduces some key technologies involved in the process of image stitching. Then the principle and characteristics of feature extraction of SIFT and SURF are analyzed based on the feature matching algorithm, and a new algorithm based on perceptual hash coincidence region detection is proposed. In this paper, a coincident region search algorithm is designed. Firstly, the overlapped region is detected to determine the effective splicing region, and then the feature points are extracted and the descriptor is used to realize the registration in the effective region. The experimental results show that the proposed algorithm can significantly improve the matching speed and efficiency, extract stable and accurate feature points, and reduce the mismatch. Compared with the existing algorithms, the proposed algorithm has better real-time performance. In the aspect of image fusion output, the problems such as obvious stitching, double shadow, uneven luminance and so on, which may appear in the registration image, are analyzed deeply. Based on the coincident region gradual coefficient fusion and the optimal stitching algorithm based on dynamic programming, the splicing slot is processed. To solve the problem of brightness difference in sequence image, the pixel distribution model of adjacent sequence image is established to equalize the brightness of image. The experimental results show that the proposed algorithm can improve the visual effect of the fusion image and obtain the ideal image output.
【学位授予单位】：沈阳理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S126;TP391.41

【相似文献】