工程车辆全景环视影像系统的研究

发布时间：2018-09-18 11:57

【摘要】：随着中国经济的快速发展,房地产市场热度不减,各项基础建设持续投入,对工程车辆的需求量也日益加大。工程车辆发挥举足轻重作用的同时,也伴随着很多安全问题。工程车辆作业环境复杂,车身庞大,仅仅依靠后视镜很难确保安全。而现代科学技术日新月异,许多工程车辆电子产品应运而生,将先进的前沿技术应用于传统的工程车行业已是必然趋势。全景环视影像系统能够实时提供车身周围的俯瞰图像,有效消除视野盲区,为驾驶者提供了十分有效的辅助作用,研究该系统并将其应用在工程车辆上具有很大的实用价值。本文分析了全景环视影像系统的背景和研究现状,深入研究了该系统的关键技术,根据工程车辆的实际特点给出了具体的实施方案。系统采用六个鱼眼摄像头对工程车辆四周进行图像采集,因采集的图像带有鱼眼畸变,因此首先将图像进行畸变矫正,然后将矫正后的图像经透视变换转换为俯视图,最后将六幅图像进行拼接融合。在鱼眼矫正算法中,本文分析了常用的几种矫正算法,并通过仿真实验进行效果对比,最终确定了使用标定法进行矫正。文中对标定原理和方法做了详细的介绍,并逐步实现了摄像头的标定,实验证明了标定法矫正能够满足系统要求。图像俯视变换阶段,本文分析了其变换模型,确定了通过寻找角点来计算透视变换矩阵的方案,并给出了参考点选取的方法。在图像匹配算法中,本文详细介绍了匹配原理和SIFT算法的实现方法,并基于传统的SIFT算法进行了研究改进,提出了基于部分特征提取的改进SIFT算法。该算法只在重合区域内提取特征点,大大减少了特征点数量和运算时间,匹配成功率和效率都有所提升。文中给出了重合区域的确定方法,通过实验对比验证了改进算法的有效性。在图像融合算法中,本文研究了常用的几种基于像素级的融合算法,通过仿真比对,选取了渐入渐出的融合算法。渐入渐出融合算法在两幅图像亮度差别较大时无法完全消除拼接缝,图像过渡不自然。本文针对此缺点对渐入渐出算法进行了改进,对重合区域的平均灰度值进行计算对比,将两幅图像灰度值调成相近后进行融合。实验证明该改进算法能有效消除拼接痕迹,使融合后图像更加自然。本文研究了全景环视影像系统各环节算法,并改进了鱼眼矫正,图像匹配和图像融合算法,提升了全景图的视觉效果,有效地减少了算法的复杂度,具有良好的应用前景和价值。
[Abstract]:With the rapid development of Chinese economy, the heat of the real estate market is not decreasing, the investment of various infrastructure projects is continuing, and the demand for engineering vehicles is increasing day by day. Construction vehicles play a pivotal role, but also accompanied by a lot of safety problems. The working environment of engineering vehicles is complex and the body is huge. It is difficult to ensure safety only by rearview mirror. With the rapid development of modern science and technology, many electronic products of engineering vehicles emerge as the times require, and it is an inevitable trend to apply the advanced frontier technology to the traditional engineering vehicle industry. The panoramic circle view image system can provide the overlooking image around the car body in real time, effectively eliminate the blind area of the field of vision, and provide a very effective auxiliary function for the driver. It is of great practical value to study the system and apply it to the engineering vehicle. In this paper, the background and research status of panoramic circle view image system are analyzed, the key technology of the system is deeply studied, and the concrete implementation scheme is given according to the actual characteristics of engineering vehicles. The system uses six fish-eye cameras to collect images around the construction vehicle. Because the collected images have fish-eye distortion, the system first corrects the distortion of the images, and then converts the corrected images into the overlooking images by perspective transformation. Finally, the six images are stitched and fused. In the fish-eye correction algorithm, this paper analyzes several commonly used correction algorithms, and through simulation experiments to compare the results, finally determine the use of calibration method to correct. In this paper, the principle and method of calibration are introduced in detail, and the calibration of camera is realized step by step. The experimental results show that the calibration method can meet the requirements of the system. In the stage of image overhead transformation, this paper analyzes its transformation model, determines the scheme of calculating perspective transformation matrix by looking for corner points, and gives the method of selecting reference points. In the image matching algorithm, this paper introduces the matching principle and the realization method of SIFT algorithm in detail, studies and improves based on the traditional SIFT algorithm, and proposes an improved SIFT algorithm based on partial feature extraction. The algorithm only extracts feature points in the coincidence region, which greatly reduces the number of feature points and operation time, and improves the matching success rate and efficiency. In this paper, the method of determining the coincidence region is given, and the effectiveness of the improved algorithm is verified by experimental comparison. In the image fusion algorithm, this paper studies several commonly used fusion algorithms based on pixel level, and selects the gradually in and out fusion algorithm through simulation comparison. When the brightness difference between the two images is large, the image transition is not natural, and the splicing seam can not be completely eliminated. In order to solve this problem, this paper improves the incremental and gradual out algorithm, calculates and compares the average gray values of the coincidence region, and then adjusts the gray values of the two images to similar values and then fuses them together. Experiments show that the improved algorithm can effectively eliminate the stitching trace and make the fusion image more natural. In this paper, the algorithms of panoramic circle image system are studied, and the algorithms of fish-eye correction, image matching and image fusion are improved, the visual effect of panoramic image is improved, and the complexity of the algorithm is reduced effectively. It has good application prospect and value.
【学位授予单位】：中国矿业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TU603;TP391.41

【参考文献】