【作者】 邓志鹏；

【导师】 岳宏； 李铁军；

【作者基本信息】 河北工业大学 ， 机械制造及其自动化， 2000， 硕士

【摘要】 目前，在工业生产中，存在着高温、核辐射、毒化气体等极限工作环境，对机器人的智能化提出了更高的要求。为了使机器人具有高度的智能，需要配备各种传感器，其中视觉传感器是最主要的。视觉系统处理的信息数据量大，这样，如何提高图象的处理速度成为许多科技工作者研究的热点课题，也是亟待解决的难题。利用4f光学系统可以完成图象傅立叶变换，实现频谱滤波。作者在此方面进行了一些研究工作，主要内容如下： 设计了光学小波变换系统，同时采用光寻址的空间光调制器（LASLM）和电寻址的空间光调制器（EASLM）来调制图象信号，LASLM将物镜的成象转为相干光的图象，此图象的读入转换过程在几十毫秒内完成。根据需要利用计算机将滤波函数输入到EASLM，使输入图象与滤波函数在频率域相乘，最后由CCD接收调制后的图象信号，进行后续处理，整个处理过程约为100毫秒，使图象的处理速度成倍的提高。 鉴于光学小波变换系统是以实现机器人视觉的路径跟踪为主要目标，并且考虑Haar小波的边界检测功能，Mexicohat小波边缘增强的作用，对整个光路信息传递进行了分析。针对LASLM的空间采样和量化，将电子学中的A/D转换去假频前置滤波器的设计应用于光路分析中，为结构设计提供理论依据。同时对量化噪声的形成进行了分析，确定为高斯噪声。 用计算机进行了仿真实验，得到了对图象的边缘提取，验证了光学系统设计的可行性。同时仿真结果证明，对于低频图象而言，选取合适的小波尺度因子，Haar小波和Mexicohat小波均可实现对图象边缘的检测和提取，并且能够有效地抑制噪声。更多还原

【Abstract】 In recent years, in industry production, there has limited working surroundings which are high temperature ~. nuclear radiation poisonous gas, it needs more requirements to robot. In order to make high intelligent for robot, we should fit various sensors on robot, the most important of all is visual sensor. The information data of robotic visual system is very large, thus, how to improve the speed of image processing has been the studying topics and is a difficult problem to be overcome. We could achieve the Fourier transform of image and filter in frequency spectrum by using 4f optical system. The author do some studying works in this domain, the main contents are the follows: The author has designed the system of optical wavelet transform, put forward using spatial light modulation of light address and spatial light modulation of electric address to modulate signals of image at one time. LASLM transform the image, which is the object image, into coherent image, this process of read-in and transforming could be finished in a few milliseconds. According to the need, we can input various functions into EASLM by using computer, and the input image multiplied by filter function in the frequency domain, at last, the signal of modulated image is incepted by Charge Coupled Device (CCD) and is processed in next steps. The time of whole processing is almost 100 milliseconds, the speed of image processing could raise a lots times. The main goal of this system based on optical wavelet transform is path tracks of robotic vision, and Haar wavelet has the edge of detection properties, Mexicohat wavelet has the edge of enhancement. Aiming at spatial sampling and quantification, the whole transfer of light path information has been analyzed. The author puts 3 forward using the design of A/D transfer antialiasing prefilters, which has been used in electronics, in the analysis of light path, thus we have theoretic foundation in the design of structure, and analyze quantification noises, which is the Gauss noise. We had gained of the image edge by using computer simulation, it proved feasibility of the optical system design. In the same time, for the low frequent image, II if we select suitable scale of Haar wavelet (or Mexicohat wavelet), we can detect the edge of image, and the noises could be effective restrain.更多还原