【作者】 张晴；

【导师】 甘棕松；

【作者基本信息】 华中科技大学 ， 计算机技术（专业学位）， 2022， 硕士

【摘要】 图像传感器作为现代光机电系统的核心部件,与其他尖端光机电设备联用可以对图像数据实时采集、处理和传输来反馈控制光机电系统。然而,使用图像传感器进行图像采集时,目前主要还是以人工调焦为主,导致调焦精度和效率较低。此外,现有的图像传感器控制软件大多只支持单一设备的控制,无法直接与其他设备进行联用,造成数据通信效率低下。针对这些问题,本文探索了自动聚焦算法中的聚焦评价函数和聚焦搜索算法两大关键技术,在此基础上开发一套集自动聚焦和设备控制于一体的图像传感器控制软件系统,并研究了自动聚焦算法对自动化、智能化及集成化图像传感器控制软件系统的优化。具体研究内容如下:(1)对聚焦评价函数进行优化研究。通过对传统的聚焦评价函数进行实验与分析,总结出目前的聚焦评价函数存在灵敏度低和鲁棒性差的问题。然后针对这些问题,提出改进的Tenengrad聚焦评价函数,并且通过仿真实验验证该函数的性能。(2)对聚焦搜索算法进行优化研究。首先对几种聚焦窗口选择方法进行分析,总结各种取窗法的优缺点和适用场景。然后对传统的聚焦搜索算法进行研究并分析它们各自存在的问题,针对算法易受局部极值点和噪声的干扰问题,提出一种粗精结合的聚焦搜索算法。最后通过实际聚焦实验测试自动聚焦算法的聚焦精度和成功率。(3)结合图像传感器的应用需求,设计并实现了一套集自动聚焦和设备控制于一体的图像传感器控制软件系统。对实验结果分析得出,改进的Tenengrad聚焦评价函数在灵敏度方面比Tenengrad函数至少提高了2.1倍,鲁棒性能比Tenengrad函数至少提升了16.76%。本文的自动聚焦算法聚焦的平均成功率达到90%,最小平均聚焦精度达到0.86μm。图像传感器控制软件系统能够实现图像传感器与运动平台设备的集成,为多种设备的集成奠定了良好的基础。更多还原

【Abstract】 As the key component of modern opto-electromechanical systems,image sensor can be used in conjunction with other advanced opto-electromechanical devices to collect,process and transmit image data in real time to feedback and control opto-electromechanical systems.However,artificial focusing is currently the main method when using an image sensor for image acquisition,which results in low focusing accuracy and efficiency.In addition,most of the image sensor control softwares only supports the control of a single device and cannot be directly combined with other devices,which results in low efficiency of data communication.In view of these issues,this thesis explores the two key technologies of focus evaluation function and focus search algorithm in auto-focusing algorithm,and develops a image sensor control software system integrating auto-focus and equipment control,and studies the optimization of auto-focusing algorithm on automatic,intelligent and integrated image sensor control software system.The main research contents are as follows:(1)Research on optimization of focus evaluation function.From the experiments and analysis of traditional focusing evaluation function,it was concluded that the current focusing evaluation function had the issues of low sensitivity and poor robustness.To address these issues,an improved tenengrad focus evaluation function was proposed,and the performance of this function was verified by simulation.(2)Research on optimization of focus search algorithm.Firstly,several focusing window methods were analyzed to obtain the performance and applicable scenarios.Then,the traditional focus search algorithms were studied and analyzed to obtain their issues.Aiming at these issues that the algorithms were easily disturbed by local extreme points and noise,a focus search algorithm combining coarse and fine was proposed.Finally,the focusing accuracy and success rate of the automatic focusing algorithm were tested by the actual focusing experiment.(3)Combined with the application requirements of the image sensor,a image sensor control software system integrating auto focus and equipment control was designed and accomplished.The analysis of the experimental results shows that the sensitivity of improved tenengrad focus evaluation function is at least 2.1 times higher than the tenengrad function,and the robust performance is at least 16.76% higher than the tenengrad function.The average success rate of the autofocus algorithm is 90%,and the minimum average focusing accuracy is 0.86 μm,which has high practical application value.The image sensor control software system can realize the integration of image sensor and motion platform equipment,and establish good foundation for the integration of various equipments.更多还原