Research On Adaptive Illumination Compensation Method For High Dynamic Range Images

Image quality is a decisive factor in image machine vision decisions.In the industrial inspection scene,due to the influence of environmental illumination,surface material of the measured object,reflectivity and three-dimensional shape,the captured image is prone to different lighting phenomenon,resulting in over-exposed areas and under-exposed areas appearing simultaneously.It is impossible to recover by means of image processing.The loss of detailed information of the acquired image affects the accuracy and reliability of industrial inspection.Based on the above problems,this paper proposes a scenario for applying industrial lighting detection using a self-applied illumination system.The main idea of this system is to adaptively fill light by actively improving the lighting conditions: reducing the light intensity in the overexposed area and enhancing the illumination in the low exposure area.This system relies on the projector-camera system to improve the quality of the acquired image through the threedimensional object surface pixel matching and based on IHDR(Iterative High Dynamic Range)algorithm.In this paper,we first use the Fourier fringe analysis method in 3D shape measurement to make every pixel in the projector-camera match on the coordinate system.The improved Fourier fringe profilometry method for eliminating zerofrequency interference problem is analyzed.An elliptic filter algorithm suitable for gray-scale Fourier profilometry is proposed.This filter can improve the accuracy of the operation.The amount of data calculation is reduced,and the recovery speed of the 3D depth information is improved.Secondly,the existing adaptive iterative algorithm is discussed and verified,and the IHDR algorithm is proposed based on the inversion iteration method and the high dynamic range algorithm.This method makes up for the shortcomings of traditional iterative algorithms with many generations and long running time.At the same time,it has better performance than HDR(High Dynamic Range)algorithm in extremely strong lighting scenes,and is suitable for reliable and real-time characteristics in industrial detection scenarios.Finally,the elliptical filter and IHDR algorithm are verified by theoretical analysis and experiment.The depth information error is below 0.1cm,the iteration number is from 10 times to 3 times,and the iteration time is about 1s?...