Automatic illumination planning for robot vision inspection system

• Published in: Neurocomputing (Amsterdam) Vol. 275; pp. 19 - 28
• Main Authors: Wang, Hesheng, Wang, Jingchuan, Chen, Weidong, Xu, Lifei
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 31.01.2018
• Subjects: Automatic illumination; Feature learning; Image quality analysis; Robot vision inspection; Automatic illumination; Feature learning; Image quality analysis; Robot vision inspection
• ISSN: 0925-2312; 1872-8286
• DOI: 10.1016/j.neucom.2017.05.015

High-quality original image is very important in robot vision inspection system and illumination is a significant component that directly affect cameras optical imaging system and plays a decisive role on image quality. To guarantee camera imaging system for high-quality images and achieve automatic illumination control in the motion of inspection robot under dark environment, this paper proposes an optimal light intensity planning method based image quality analysis. It is mainly achieved by building a computational model to automatically predict optimal light intensity values for desired image quality when camera observation distances fluctuate. Before regression modeling, it is necessary to extract discriminative features representing image quality. We design feature extractor by deep learning instead of human engineers which required careful engineering and considerable domain expertise. Deep learning methods are representation-learning methods that allows a machine to be fed with raw data and to automatically discover the representations needed for regression or classification. Experimental results demonstrate the feasibility and efficiency of this method.