摘 要
为实施国际航协提出的“简化商务、便捷出行”的核心计划，研发自助行李托运系统已经成为重要的解决方案。其中，行李信息的检测包括行李重量的采集、规格的检测以及行李标签的扫描。针对行李规格的自动检测问题，研究了基于双目立体视觉的检测算法，以实现自助行李托运系统中对行李的非接触检测。
首先，基于自助行李托运系统特殊的硬件环境，设计了平行双目结构的行李图像采集系统，使系统能够在有限的空间内采集行李的立体图像对。设计了一种基于棋盘平面的系统标定法，获取摄像机的内外参数。
其次，采用自适应最优阈值法对行李图像进行预处理并提取行李特征。设计了一种能够描述行李几何形状和轮廓的二维canny边缘检测算法。在外极线、唯一性、灰度相容性等基本约束和视差梯度附加约束条件下，设计了基于特征的区域灰度匹配方法，采用双向匹配策略，减少误匹配，提高了匹配的可靠性和稳定性。
最后，考虑了双目视觉系统中空间点三维信息恢复方法，包括最小二乘法、中点法以及三角法。采用三角法并结合摄像机标定参数计算行李轮廓的三维信息。
基于上述关键技术和方法，对几类行李进行检测实验，验证了本文算法的可行性和有效性，具有一定的应用价值。
关键词：行李规格检测；立体视觉；特征提取；立体匹配；三维信息恢复
Abstract
In order to implement fully “simplify business” program put forward by IATA(international air transport association) and shorten the boarding time, researching the self-service luggage system has become an important solution. Passengers can use it to make the identification and print boarding pass and baggage tag. The system can automatically detect baggage information, and finally complete luggage checked. Among them, the baggage information detection includes baggage weight collection, specifications detection and luggage label scanning. For the luggage size automatic detection problem, a set of detection algorithm based on binocular stereo vision is proposed to realize non-contact detection on the self-service luggage system.
Firstlyl, according to the hardware environment of self-service luggage system and binocular stereo vision structure model. It designs luggage image acquisition hardware system of the parallel binocular structure. At the same time, the camera imaging model and calibration method were studied, using a system calibration method based on the chessboard plane to get the camera parameters.
Secondly, using the optimal threshold value method to preprocess the images. Then, in feature extraction, designing 2D canny edge detection algorithm to describe the luggage geometric shape. After that, According to basic constraints and disparity gradient additional constraint conditions, designing the regional gray matching method based on feature, and using the two-way matching strategy, reducing the error matching, It ensures the reliability and stability of the matching.
Finally, researching the calculation methods of space point 3D information of the binocular vision system. According to the parallel binocular structure, using the triangular method to calculate 3d information of the baggage.
Based on the technology and method mentioned above, testing the shape of different size luggage can realize the non-contact detection of baggage specifications.
Key Words: Luggage Specifications Detection; Stereo Vision; Edge Feature; Stereo Matching
目 录
