Automatically detecting surface defects from images is an essential capability in manufacturing applications. Traditional image processing techniques were useful in solving a specific class of problems. However, these techniques were unable to handle noise, variations in lighting conditions, and background with complex textures. Increasingly deep learning is being explored to automate defect detection. This survey paper presents three different ways of classifying various efforts. These are based on defect detection context, learning techniques, and defect localization and classification method. The existing literature is classified using this methodology. The paper also identifies future research directions based on the trends in the deep learning area.
In many manufacturing applications, robotic manipulators need to operate in cluttered environments. quickly finding high quality paths is very important in such applications. This paper presents a point-to-point path planning framework for manipulators operating in cluttered environments. It facilitates finding a balance between path quality and planning time. The framework dynamically switches between various strategies to produce high-quality paths quickly. In this work, (1) we extend a previously developed sampling-based modular tree-search, (2) we add new strategies and scheduling logic that decreases the failure rate as well as the planning time compared to our prior work, (3) we also present theoretical reasoning behind strategy switching and how it can help decrease planning times and increase path quality. Specifically, we present a strategy that can sample effectively in challenging regions of the search-space by using local approximations of the configuration space. We also present an inter-tree connection strategy that adapts to collision information gathered over time. We introduce a scheduling rule that regulates the exploitation of focusing hints derived from the workspace obstacles. Together, these new extensions the reduce average failure rate by a factor of 4 and improve the average planning time by 22% over previous work.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.