Research on Intelligent Inspection of Highway Construction Quality Integrating Field Image and Specification Texts

Construction inspection is critical to ensuring the quality and long-term performance of highway infrastructure. In the current practice the inspectors need to manually gather and interpret the lengthy quality standards for inspecting and evaluating the highway construction work, which is subjective, error-prone, and time-consuming. Many transportation agencies have developed inspection forms to organize construction requirements that reside in textual documents (e.g., standard specifications, construction inspection handbooks, and quality standards) in the format of checklists in order to reduce the workload for inspectors and enhance productivity. However, due to the missing link between the inspection forms/checklists and the construction work under inspection, the inspectors might need extra work to find the applicable forms/checklists for inspection. This paper thus proposes an image-based approach to establish the missing link. First, use computer vision algorithm to understand the construction scene in the image and deduce the construction activity through the elements in the scene. Second, the semantic structure underlying the quality standards for highway construction (a Chinese quality standard for highway engineering is selected as the case) is analyzed, followed by the re-structuring of the quality requirements as inspection forms. Then the inspection forms are associated with their relevant construction activities. Last, an image-enabled prototype is presented to illustrate the generation of customized inspection forms for a given construction site image. With this newly developed tool, field inspectors can get rid of the overwhelming texts in the quality standards and can be equipped with the necessary knowledge regarding what and how to inspect.