|英文摘要: ||Auto focus is a key technology in new electro-optical tracking and detecting system. In order to improve performance of electro-optical tracking and detecting products, this dissertation fulfilled some important theoretical and technical studies about autofocusing, such as auto focus theories and methods, focus function and algorithm, focus framing technologies, we also developed some effective methods to control zoom-lens automatically. This dissertation achieved a numbers of outcomes.
Basic theories and methods of autofocus based on image processing are studied in detail. By some experiments, we analyzed the capability of frequency spectrum function, gradient function and entropy function used in autofocus. We also have studied autofocus methods for moving target, discussed the Characteristics of autofocus for moving target and influence to smoothness of focus measure curve and proposed the precondition for moving target’s autofocus and calculatingly methods .
A new autofocus evaluate function based on the sum of order grads is proposed. Traditional function based on grads should calculate all grads in image area. If object or background is changed, the sum of most small value grads would make the lens be drive to wrong position. The new focus function only calculate some grads which sorted ahead, the feasibility of this algorithm was proved by a series of focus experiments when size of object changing in a certainly range. To reduce the disadvantage caused by pulse noises, another function based on the number of grads above threshold is proposed. This function compares the number of grads above a dynamic threshold in current frame image with that in last frame, and gets the direction of autofocus. Curve of experiments result proves this function has excellent veracity and reliability.
By analyzing the flow of autofocus, we have concluded that judging current image be in focus or not is necessary before doing autofocus. We proved the corresponding connection between degrees of definition and edge acutance by means of experiments. So, a judgment algorithm based on the edge feature was proposed, which used for judging current image be in focus or not. It can express the degree of definition by math parameters and has good performance in judging the images of sky objects.
Appropriate focus window selection is an important factor in the process of autofocus. Most of pixels in focus window should be from object image, thus can reduce the disturbing effect cause by pixel from background. According to the feature of image, using technology of aim tracking for reference, we proposed two schemes about focus window selection: based on tracking object’s center of mass and based on correlation match tracking method. The first method is suitable for situation that background gray scale is smoothly. It has advantages of rapid reaction and steady tracking object. The second method has power capability of anti-disturbance, which suitable for situation that image in low SNR and low contrast.
The theory of changing focus based on image processing is studied and the flow of changing focus is proposed. How to get the real size of object in image is the key technology. After discussing the advantage and disadvantage of image segment by single threshold and by adapting threshold, we presented an algorithm of image threshold segment based on edge grads. It use grays located in edge transition region as threshold and reduce the influence caused by background and noise.
An autofocus and auto-zoom system based on PC has been designed. After performing a series of autofocus and auto-zoom experiments, we conclude this system has excellent veracity and reliability.|
|Appears in Collections:||光电技术研究所博硕士论文_学位论文|
|File Name/ File Size
倪军. 基于图像处理的自动调焦变焦技术研究[D]. 光电技术研究所. 中国科学院光电技术研究所. 2007.