Collaborative Research: Using Computer Vision to Measure Neighborhood Variables Affecting Health

该项目将开发一种自动化方法，利用计算机科学的进步，系统地观察和记录大规模邻近环境的物理条件。 邻里环境在塑造个人和社区的健康方面发挥着重要作用，从而导致美国的不平等。过去的研究表明，社区中存在身体障碍、维护不善的财产和空地会对身心健康产生负面影响，吸引更多的犯罪和混乱，并导致社区撤资。 这项研究所产生的措施将有助于审查这些进程，并为科学研究界提供一个强大的资源。 通过让政策制定者、从业者和公众跟踪社区的进展和目标改善情况，它们也将带来更广泛的好处。该项目利用谷歌街景图像--最大的公开可用的街区视觉外观纵向数据集--并将利用亚马逊的土耳其机器人--一个众包平台--和现有的实地调查数据来识别身体障碍和维护的指标，如垃圾和破旧的建筑物，在三个不同城市的街道段样本上：波士顿，底特律和洛杉矶。这些数据将用于训练一种算法，该算法利用了机器学习和计算机视觉的最新进展。将在整个过程的每一步中测试用于确定特征和措施的方法的可靠性和有效性。由此产生的邻里物质条件的纵向测量将与在三个城市中的每一个进行的纵向健康调查相联系，以分析邻里物质条件与健康之间的关系。此外，新的测量方法将作为多个城市物理社区条件纵向测量的公开数据库发布。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project will develop an automated method, using advances in computer science, to observe and record systematically the physical conditions of neighborhood environments at a large-scale. Neighborhood environments play a significant role in shaping the health of individuals and communities, consequently contributing to inequality in the U.S. Past research suggests that the presence of physical disorder, poorly maintained properties, and vacant lots in neighborhoods can negatively affect physical and mental health, attract more crime and disorder, and lead to neighborhood disinvestment. The measures resulting from this study will facilitate examination of these processes and provide a powerful resource for the scientific research community. They will be more broadly beneficial as well by allowing policymakers, practitioners, and the public to track neighborhood progress and target improvements.The project takes advantage of Google Street View imagery--the largest publicly available longitudinal dataset of visual appearance of street blocks--and will use Amazon's Mechanical Turk--a crowdsourcing platform--and existing field survey data to identify indicators of physical disorder and maintenance, such as trash and blighted buildings, on a sample of street segments across three distinct cities: Boston, Detroit, and Los Angeles. These data will be used to train an algorithm that draws on recent advances in machine learning and computer vision. Reliability and validity of the method for identifying characteristics and measures will be tested throughout each step of the process. The resulting longitudinal measures of the physical conditions of neighborhoods will be linked to longitudinal health surveys conducted in each of the three cities to analyze the relationship between physical neighborhood conditions and health. In addition, the new measures will be released as a publicly available database of longitudinal measures of physical neighborhood conditions across multiple cities.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.