Research On Fast Auto-focus Technology Of Short Focus Depth Optical Camera Based On Image Features

With the development of space optical camera,the requirement for the imaging resolution and imaging range are gradually improving.Hence,the space camera with both higher resolution and wider field of view has a better prospect for development.Due to the limitation of imaging performance parameters,the space optical camera with large aperture and large field of view has a shorter focal depth.However,for the space optical system with a shorter focal depth,there easily appear random defocus error,thus resulting in worse imaging quality.This is caused by the shake of space launch and changes in environmental factors,such as temperature,pressure,etc.,during the running process.Therefore,automatic focusing technology is a key technology to ensure clear imaging.Compared with the traditional self-collimating focusing method and the focusing method with additional sensor for the space camera,the automatic focusing method based on digital image processing is more convenient and flexible.But for the remote sensing image target,the existing auto-focus method applied to other imaging systems still has slow focusing speed and poor anti-noise interference ability.Therefore,according to the characteristics of remote sensing images,this thesis studies and improves the focusing evaluation function,focusing search strategy and depth from defocus method based on the defocusing amount evaluation.Firstly,the imaging model and defocus principle of the optical imaging system are introduced,as well as the classical focusing evaluation function is introduced by category.Owing to the deficiency of traditional algorithm for the sensing image,the focusing evaluation function based on feature region is put forward.The influences of invalid pixels and noise on the accurate focusing are decreased by selecting regions with rich information.Simultaneously,some problems are solved,such as the decrease in insensitivity of evaluation function curve under great effect of noise,the occurrence of local extremum,and the blindness of focusing window selection for remote sensing image.Then,even under the influence of a large amount of Gaussian noise with a variance of 0.01,the evaluation function curve has a high sensitivity.Secondly,in view of more search steps at early stage for depth from focus method with a large defocusing amount and the occurrence of local extremum,the depth from defocus method based on remote sensing image is studied to accelerate the focusing speed and improve the focusing accuracy.Moreover,the spatial-domain convolution/deconvolution transform methods are used to evaluate the defocusing amount by two images.And the algorithm is further improved so as to make the mean of estimation error less than 10% under the effect of random noise 11.5d B within the defocus range of 17 times the focal depth.Finally,based on the characteristics of high focusing accuracy of short focal depth,focusing imaging detector for large area detector array splicing can be adjusted more rapidly and accurately.Thus,the implementation system is constructed,and the focusing accuracy for the algorithm is verified.It is found that there would only appear20～40 um focusing error with a small probability in many cases.Furthermore,the depth from defocus method is verified.In most cases,the evaluation error of the remote sensing image target by actual shoot is less than 15%,with a better effect.In combination with depth from focus method,the precise focusing can be accomplished by searching in a small range,and then the whole focusing process can be accelerated.