Research On 3D Measurement Technology Of Surface Structure Light Based On Molar Fringe And Neural Network

Optical 3D measurement technology is widely used in manufacturing,industrial robots,warehouse logistics,and autonomous vehicles because of its non-contact,high precision,large area,etc.This huge application demand has also made the world more and more research teams and personnel have joined in the scientific research and exploration.The structured light three-dimensional measurement technology has gradually become the mainstream method in the field of industrial measurement by its high precision,high speed and flexibility.The high-precision,real-time optical measurement system is the purpose of industrial three-dimensional measurement technology research.This article is based on monocular surface structured light measurement technology.Used the current mainstream phase shift method and multi-frequency heterodyne method to study the related aspects of structured light measurement.Firstly,the theoretical research and experimental results are used to study the causes of the phase failure decoded in the structured light system under the conditions of high and low darkness.It is found that the phase series calculation in the multi-frequency heterodyne phase is affected by the fringe pitch multiplier amplification error jump,and derive the corresponding error jump range,quantitatively determine the effect of noise on the phase decoding.Based on the principle of error and the principle of fringe interference,a new phase-dissolving method is proposed based on the Molar sequence fringe.This method effectively preserves the high-frequency and high-precision of the traditional phase decoding method,and fundamentally avoids the process of phase error transmission and amplification during the phase calculation.Experiments and simulations show that the new algorithm effectively improves the phase decoding process under high and low dark conditions and improves the accuracy of phase.Secondly,this paper proposes a phase-coordinate mapping relationship based on BP neural network.Using the powerful function of neural network to fit,generalize,and parallel computing to improve the 3D reconstruction speed of high-precision structured light.At the same time,this paper proposes a flexible high-precision training set acquisition method,which makes the calibration and training of the system more flexible.The experimental results show that the real-time performance of phase-coordinate calculation based on BP neural network is about 100 times higher than the traditional method.Finally,we combined the two methods proposed in this paper to measure a standard plane and perform absolute accuracy analysis.The final absolute deviation is less than 0.03 mm.At the same time,we measured a the complex surface which is difficulty measurement by the traditional method and compared with the traditional method results.The final result shows that the new method retains the high precision,and greatly improves the stability and real-time of the three-dimensional reconstruction of the surface structure light,and has a good wide applicability.