Research Of 3D Reconstruction And Its Key Technologies Based On Binocular And Linear Structured Light

3D reconstruction is an interdisciplinary technology of multi-disciplinary such as computer image processing,optics,control and hardware.It is a high-precision,high-efficiency and non-contact measurement technology,which can meet diversified measurement needs and plays an important role in the fields of industrial detection,reverse engineering,biomedicine and virtual reality.Moreover,combined with artificial intelligence technology,it can achieve information mining,which has important scientific research value.However,compared with developed countries,China's 3D reconstruction technology started late,and the reconstruction accuracy and application are urgent to be improved.Based on this,this paper takes high-precision industrial detection and other projects as the starting point to carry out relevant research on the three-dimensional reconstruction technology based on "binocular camera + line structured light".At present,the technology has the following problems:(1)Algorithm: the structure parameters of binocular reconstruction system has a direct impact on the reconstruction accuracy,and it is lack of systematic theoretical analysis;It affects the reconstruction accuracy a lot due to the low calibration accuracy of the camera and the accumulated error produced in the reconstruction process;(2)Applicability:because of the lack of effective control of lighting source and laser power,it makes the landmark of the laser line extraction accuracy is not high;(3)Data completeness:the 3D reconstruction system usually only outputs colorless point clouds,which causes the lack of target color information;(4)Degree of automation: lack of design and analysis of automatic 3D reconstruction system.The above four is also the aim of this paper.Facing the above problems,the research work of this paper includes the following four aspects:1.The analysis model of structural parameters and measurement error of binocular structured light three-dimensional reconstruction system is established,and the influence of structural parameters on error propagation coefficient is emphatically analyzed.The results show that,when the included angle between binocular optical axis and baseline is between 30 degrees ? 50 degrees,when the field angle is near 0degrees,and when the ratio of baseline distance to depth L/Z is between 0.6?3.0,the system reconstruction error is relatively small,and the experimental results are consistent with the theoretical analysis.A coordinated optimization method of calibration parameters based on re projection error is proposed.The experimental results show that,when the displacement of the mark point is from 30 mm to 210 mm,the position error of reconstruction using the calibration parameters before optimization is 0.03 ? 0.7mm.After optimizing the calibration parameters,the reconstruction position error is 0.02 ? 0.15 mm,and the reconstruction accuracy after optimization is significantly higher than that before optimization.Based on the above research results,a binocular reconstruction system is built and the standard ceramic ball is reconstructed.The fitting diameter error is 0.022 mm,which is lower than the measurement accuracy of other binocular reconstruction systems in China;The two fixed surfaces of the right angle block are reconstructed,and the deviation between the fitting angle and the measured value of the electronic angle ruler is 0.027 degrees.The measured results show that,the binocular reconstruction system studied in this paper has higher measurement accuracy and can meet most of the needs in industrial detection and measurement.2.The brightness control method of illumination source and laser in binocular structured light reconstruction system is proposed.Lighting source control method:adjust the brightness of the lighting source based on the effective recognition rate and saturation rate of the mark point.The experiment shows that this method can adjust the brightness of the lighting source in real time according to the recognition of the mark point,which is helpful to improve the adaptability of the reconstruction system to the ambient light.Laser control method: adjust the laser brightness in real time according to the effective recognition rate and saturation rate of the laser line.The experiment shows that,this method can adjust the brightness of the laser in real time according to the recognition of the laser line,which is helpful to improve the adaptability of the reconstruction system to complex surfaces.Compared with the traditional manual adjustment method,the above two adjustment methods not only expand the application scope of the reconstruction system,but also improve the reconstruction efficiency.3.A three-dimensional point cloud color restoration method is proposed: Taking the coded mark points as the positioning medium,using the colorless point cloud and color images of the target,the real color three-dimensional point cloud model of the target surface is established.Firstly,the binocular reconstruction system is used to obtain the world coordinates of the target's colorless point cloud and coding point;Secondly,the color images of the target and coding points are collected by the color camera,and the high-precision image coordinates of the coded points are obtained by using Canny edge detection and sub-pixel positioning algorithm;Thirdly,the improved direct linear transformation(DLT)method is used to solve the pose matrix of the color camera relative to the world coordinate system,and then the mapping relationship between the point cloud and the color image pixel is established;Finally,the interpolation method is used to calculate the color of the local point cloud and complete the color fusion of the whole point cloud.In the experiment,the color three-dimensional point cloud of two targets is completed.The color spatial error is less than 0.6mm,and the calculation efficiency reaches 73000 points / s.Through visual observation,the color reality is very good.This method realizes the color restoration of point cloud,and its accuracy and efficiency meet the usage requirement.4.Based on the development of clothing personalized design system,the automatic three-dimensional reconstruction system is designed and implemented.According to the measurement index of the system,the overall design scheme of "double guide rail + four groups of binocular detectors" is put forward;Based on the error analysis model,the optical design and accuracy analysis of the binocular detector are completed.The reconstruction accuracy of the binocular detector is0.09mm;The transmission mechanism and fine-tuning mechanism in the automatic three-dimensional reconstruction system are designed in detail;The overall calibration scheme of the system is designed and the overall calibration of the system is completed;In the prototype experiment,the deviation of the characteristic parameters extracted from the automatic reconstruction system relative to the measurement results of the portable reconstruction system is less than 1mm.The point cloud computing efficiency of the automatic reconstruction method is 265500 points/s,while the portable reconstruction efficiency is up to 200000 points/s.Therefore,the reconstruction accuracy and efficiency indicators of the automatic reconstruction system meet the expected objectives.In this paper,the binocular structured light three-dimensional reconstruction system and its key technologies are deeply studied.In the meantime,a series of three-dimensional reconstruction equipment and software are developed.The reconstruction effect is very good after being used by users such as the Ministry of public security,Guangdong Key Laboratory,Changchun Railway Bus Company,engine R ? D enterprises and the First Hospital of Jilin University.