【作者】 李沛；

【导师】 吕海宝；

【作者基本信息】 国防科学技术大学 ， 仪器科学与技术， 2009， 博士

【摘要】 尾旋运动是飞机最复杂也是最危险的一种极限飞行状态,是造成飞机失事的主要原因之一。立式风洞尾旋试验作为研究尾旋问题的主要手段,是为了分析和预测尾旋,以防止飞机实际尾旋事故的发生。目前我国已经具备开展型号自由飞尾旋试验的能力,但模型运动参数传统测量方法耗时长,成本高。为此课题组研制了某立式风洞配套的具有自主知识产权的飞机尾旋运动测量系统。本文在总装高新技术专项“某立式风洞建设”的支持下,围绕飞机尾旋运动测量中的关键技术问题开展深入研究,主要包括基于立体视觉的模型姿态测量方法研究、基于人工标识的立体匹配问题研究、基于滚球法的不规则直线段识别问题研究、基于信用度分类的立体视觉遮挡问题研究以及立式风洞飞机尾旋运动测量系统研制等。论文的主要研究内容和创新点归纳如下:1、针对立式风洞飞机尾旋运动姿态参数测量要求,提出了一种基于立体视觉技术的非接触测量方法,通过立体图像对恢复空间物点三维信息,进而由空间物点解读模型运动参数。该方法相对传统的测量方法具有试验效率高、数据密度大等优点,能够实现自动化和智能化测量。2、针对立体视觉测量技术应用中的立体匹配这一难题,提出了基于人工标识分类匹配的解决方法。该方法通过对人工标识的独立识别和标识位置的图像分类实现左右图像对立体匹配,探讨了标识尺寸、数量和分布等属性构造原则,研究了基于标识质心的中心点定位。3、尾旋试验图像中的人工标志臂表现为具有形变的不规则直线段,难以用现有直线段检测方法进行识别。本文首次提出了一种基于滚球法的检测技术,用于不规则直线段检测。该方法通过检测窗口沿直线段滚动的轨迹提取直线段延伸方向,由检测窗口移动参数识别直线段。试验表明该方法能有效地识别不规则直线段,对线段的宽度形变有较强的容错调节能力,已成功应用于十字、T字形等人工标识的图像识别中。4、遮挡问题是立体视觉技术应用中的又一难题,常用解决方法需要恢复连续的被遮挡空间,计算量大且工程实现难度高。为此,本文提出了一种基于信用度分类的解决方法。该方法通过标识点间的距离约束在被遮挡空间搜索目标点的最佳近似点,并引入信用度因子的加权计算,由高信用度的辅助点恢复被遮挡点。试验证明,该方法能有效恢复被遮挡点,使尾旋试验图像的可计算率由60%提高到80%以上。5、成功研制了拥有自主知识产权的立式风洞飞机尾旋运动测量系统。地面定参数旋转天平试验、实际飞机模型尾旋试验以及系统误差综合分析结果表明该系统在测量视场范围3m×3m至6m×6m情况下俯仰角、滚转角等姿态参数测量精度优于0.5%;旋转角速度相对测量精度优于0.96%。立式风洞飞机尾旋运动参数测量系统于2007年12月通过了项目鉴定,鉴定专家一致认为系统总体性能指标与国际上风洞技术发达国家的同类系统相比,明显优于俄罗斯,处于国际先进水平。目前该测量系统已成功应用于多个重点型号的尾旋自由飞试验以及空军跳伞训练辅助测试等。更多还原

【Abstract】 Spin is the very dangerous flight status of aeroplane and has caused many air disasters. The free spin test in vertical wind tunnel is an important way to study spin and is very useful for spin avoid. Conventional methods to estimate the plane attitudes in spin are time consuming, inefficient and uneliable. To develop efficient system for spin test, supported by the military project‘Vertical Wind Tunnel Build’, this dissertation concentrates its attention on the key technologies of estimating the plane attitudes in the free spin test efficiently. With the purpose of developing the plane attitudes estimation equipment with own intellectual property rights, the dissertation studies several techniques, including the plane attitudes estimation techniques based on stereo vision, the techniques to design the manmade vision signs, the improved line detection algorithm adapted for amoeboid line segments recognition, the occlusion in stereo vision and the construction of the stereo vision equipment to estimate the plane attitudes in spin time.The main contents and contributions of this dissertation are as follow:1. Aiming at estimating the plane attitudes in spin time efficiently with advanced techniques, a method is proposed based on the stereo vision technique and a set of manmade signs. The stereo vision technique should conceive the 3D coordinates of manmade signs and from their coordinates, the spin parameters of model should been worked out. Compared with traditional ways, this method is automatic and intellectualized with advantages in better efficiency and heigher data density.2. Matching is a difficult problem for stereo vision applications. To solve the matching problem, a method though a set of manmade signs is set up. The matching way is based on the absolute distinguish and classification of manmade signs. The signs’design principle is shown such as their size, amount and distribution. Finally, some studies focus on the signs classification and orientation of the signs’center.3. In spin tests images, manmade signs are represented as irregularly line segments and are hardly distinguished with existing line detectors. To solve the problem, a new way for blurry line segment detection is presented, named rolling ball detection. In this way, a detecting window is just like a ball rolling through the line segments. The rolling ball method does a good work to detect the blurry line segments and is tolerant of segments’wide, length and even deformations. The questions, how detecting window steps over the intersection area, are also studied, so it is feasible to detect the manmade signs. The new line detector is also useful to distinguish abnormal segments for other works.4. Occlusion is another hard problem in stereo vision. The previously proposed means always count out all the occluded space. It is computationally expensive and hard to come true. In this dissertation, a new way based on the credit coefficients is presented. By the distance constraint condition, the occluded signs would be searched until the best optimal estimates are gotten. The credit coefficients are introduced into the process to improve the accuracy of positions. Effects of the assistant signs are also analyzed and to be sure that the occluded signs are searched based on the very reliable signs. The credit coefficients could whittle the affections of signs with poor definition. Simulations and experiments on real images validate our method useful and the spin figures using probability has been increased for 60% to above 80% with occlusion problem solved.5. A system used to estimate the plane attitudes in spin time in the vertical wind tunnel is built up successfully with own intellectual property rights. The fixed parameters rotary balance experiments on ground and the actual spin tests have validated the system meet all targets. In a vision field from 3m×3m to 6m×6m, the relative precisions of angles are better than 0.5% and the relative precision of spin rolling velocity is better than 0.96%.In Dec.2007, the system has passed the item identification with the comments that its performance index is better than traditional system in Russia and has come up to an international advanced level. The system has been used for several aircraft types for spin tests and has assisted to airman parachuting trainings.更多还原