Research On Some Key Technologies In Automatic Focusing System Based On Image Processing

Image is an important carrier for information transmission.How to obtain a high-quality image has always been the direction and goal of human exploration.Image quality is directly related to the subsequent image processing and application.As one of the key technologies of optical imaging system,auto-focus technology is widely used in satellite navigation,power line patrol,robot vision,automatic monitoring and other fields.With the popularity of various miniaturized imaging devices such as digital cameras and smart phones,consumers have higher and higher requirements for image quality.In order to improve the image quality,two important parts of image acquisition system is studied,including auto-focusing and image screening,as well as their application.Firstly,aiming at the problem that the sharpness evaluation function in the AF system based on the depth of focus method is sensitive to noise,a new image sharpness evaluation function is proposed.Firstly,the gradient accumulation calculation is carried out by using four-directional Scharr operator to realize the selection of double region focusing windows;Secondly,the focusing windows are decomposed by two-dimensional empirical mode,and the first two components containing more image details are extracted;Finally,the energy gradient value is calculated by weighting the two components to obtain the definition evaluation function value.At the same time,in order to realize the quantitative analysis of sharpness evaluation function,a sensitivity evaluation index in fine focusing stage is proposed.The experimental results show that compared with the common sharpness evaluation function,the proposed method has the greatest sensitivity and the strongest anti-noise ability.Secondly,aiming at the problem of image quality assessment,this paper proposes two blurred image quality assessment methods.For learning-free based method,inspired by the theory of auto-focusing based on image processing and maximum local information,this paper proposes a no-reference blurred image quality assessment method based on double maximum local information.Firstly,the maximum gradient and local entropy are combined to extract the region of interest,and then the gradient information from different color channels is obtained to drive the final evaluation value.In the aspect of learning-based method,considering that machine learning method has stronger feature learning ability and better performance,this paper proposes a structure of structure feature-based no-reference blurred image quality assessment method.In the time domain,a new weighted local binary model is proposed to extract features from multi-resolution images using maximum local change mapping.In the frequency domain,the entropy information and gradient information of multi-scale log Gabor filtered image are combined to extract features.Then,the features are mapped into the quality score by support vector regression.The experimental results show that compared with some of the most state-of-the-art methods,the proposed methods have better evaluation results and are more in line with the human visual system.Finally,this paper takes Arduino MCU as the core to build a simple machine vision system.The sharpness evaluation function proposed in this paper and several commonly used functions are tested on the system for focusing accuracy and real-time performance evaluation.The experimental results show that the proposed focusing method is feasible and has the best focusing accuracy.At the same time,for the acquired images,the corresponding datasets are established,and two proposed image quality assessment methods are verified.The experimental results show that the proposed method can better screen out clear / blurred images.