CAREER: A Robust, Low-Cost Vision System for Assessing Surface Appearance

这项研究的一个长期目标是开发一种低成本、健壮、自动化的在线监控系统，该系统部分基于感知的表面外观来控制汽车涂装过程。作为重要的第一步，这项职业研究计划将：(1)开发以反射模型为合理基础的视觉系统硬件和软件设计；(2)确定最佳图像属性和处理技术，以利用该设计与感知表面外观之间的联系；以及(3)构建基于捕获图像的原型分类方案来评估表面质量。与现有方法并行使用小规模实验、模拟、原型测试和现场测试的四管齐下的方法，预计将产生一种照明设计和传感协议，该协议还将为当前的评估方法、系统软件和线路操作员都容易解释的捕获图像(和派生属性)产生改进的结果，以及一种低成本、灵活的设计，可以用商业上可用的组件来构建。对汽车涂料的性能要求是相当高的。除了提供保护和增强耐用性外，涂层还必须具有视觉吸引力。众所周知，油漆表面的外观极大地影响客户对该产品质量的看法，汽车制造商每年花费大量资金修复与油漆有关的问题。因此，在线监测表面质量的可靠方法是至关重要的。由于人们对汽车普遍着迷，学生们直观地理解油漆工作的质量是如何激发这种兴奋的，这项研究计划很容易整合到各种教育活动中。作为这一努力的一部分，预计的具体教育和外联目标包括：(1)一个多样化的研究小组，向本科生、K-12学生及其教师提供外联和指导；(2)顶峰设计项目；(3)现有本科课程的研究数据分析(以及随后研究成果的传播)。教育计划的预期好处包括为不同人群提供协作性、动手实践的研究经验和指导，以及增强首席研究人员建立和维持具有竞争力的研究计划的能力，以吸引准备充分的、不同的学生群体。

A long-range goal of this research is the development of a low-cost, robust, automated on-line monitoring system that controls the automotive painting process based, in part, upon perceived surface appearance. As a significant first step, this CAREER research plan will: (1) develop a vision system hardware and software design which has reflectance models as a rational basis; (2) determine the best image attributes and processing techniques to exploit the link between this design and perceived surface appearance; and (3) construct a prototype classification scheme to assess surface quality based on captured images. The four-prong approach, which employs small-scale experiments, simulation, prototype testing and on-site testing in parallel with existing methods, is expected to yield an illumination design and sensing protocol which also yields improved results for current evaluation methods, captured images (and derived attributes) which are readily interpreted by both system software and line operators and a low-cost, flexible design which can be built with commercially available components. Performance demands for automotive coatings are considerable. In addition to providing protection and enhancing durability, coatings also have to be visually appealing. It is well established that the appearance of a painted surface greatly affects a customer's perception of that product's quality and automotive manufacturers spend considerable sums each year to repair paint-related problems. Therefore, robust methods to monitor surface quality on-line are critical.Since there is a widespread fascination with automobiles and students intuitively understand how the quality of the paint job fuels that excitement, this research plan is very easily integrated into a variety of educational activities. Specific educational and outreach objectives projected as part of this effort include (1) a diverse research group that provides outreach and mentoring to undergraduates, K-12 students and their teachers; (2) capstone design projects; and (3) the analysis of research data (and subsequent dissemination of research results) in existing undergraduate courses. Expected benefits of the educational plan include collaborative, hands-on research experiences and mentoring for a diverse population, and enhancement of the principal investigator's ability to build and maintain a competitive research program which attracts a well-prepared, diverse group of students.