Research on monocular intelligent depth measurement method based on liquid bionic vision system. (15th March 2023)

• Record Type: Journal Article
• Authors: Lv, Jianming; Gan, Zihao; Hong, Huajie; Yan, Xiangda; Sun, Zhaomei
• Is Part Of: Measurement. Volume 209(2023)
• Publication Date: 2023-03-15
• Subjects: Depth measurement -- Structure optimization -- Neural networks -- NSGA-II -- Liquid lens calibration; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2023.112496
• Languages: English
• ISSNs: 0263-2241

Abstract

Abstract: Monocular depth measurement is an important fundamental technology in machine vision with a wide range of application scenarios, while traditional monocular focus measurement systems have deficiencies such as large size and slow response time. Low measurement accuracy of existing monocular measurement algorithms and can only measure for specific targets. Based on this, A liquid bionic vision system is proposed and a monocular depth measurement model is developed, the structural parameters are optimized using the NSGA-II algorithm. After calibrating the liquid lens, a neural network-based monocular intelligent depth measurement method is proposed, expressing the relationship between the liquid lens control current and the target depth. The experimental results show that the proposed method and the designed system perform well in terms of measurement accuracy, stability and applicability, with a relative error of less than 1.11 % within a measurement range of 1 m and single measurement time does not exceed 410 ms. Highlights: A monocular depth measurement model based on a liquid vision system is proposed. The liquid bionic vision system was optimally designed using the NSGA-II algorithm. A neural network-based intelligent depth measurement method was designed. The proposed measurement method has excellent accuracy and generalization ability.