嵌入式黑白色选系统研究

发布时间：2018-06-14 09:25

本文选题：机器视觉 + 嵌入式系统　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：基于机器视觉的色选系统是一种智能化的筛选设备,一般涉及到光学成像、图像处理和电信号控制等方面。目前,色选系统中的图像信号处理过程多应用计算机作为控制核心,存在体积大、筛选效率低等弊端。如今,嵌入式技术的跨越式发展为色选系统提供了新的思路和方法,而黑白色选设备以其结构简单、性价比高等优势得到了广泛的应用。因此本文研究基于机器视觉的嵌入式黑白色选系统,以FPGA为控制核心,具有实时性好、可靠性高的特点,在工农业等领域具有广阔的应用前景。本文以黑白线阵CCD相机为传感元件,以Xilinx公司的Spartan-3E系列的XC3S500E芯片为处理核心,设计并制作FPGA主控电路板和输出驱动电路板。利用FPGA的IP核技术在其内部建立双口RAM,实现物料信息的读取和存储的分离控制。由于光照不均现象会给测量结果带来偏差,因此采用VHDL硬件描述语言编程实现视觉数据补正功能,对像素灰度值进行非线性补偿,再编程实现比较判断以及延时输出驱动信号等功能。驱动电路通过对气阀的开通与关断进行控制,实现对残次品的剔除。以稻米检测为研究对象,应用MATLAB仿真工具对多种图像处理算法进行研究,并对其进行改进,使其适用于FPGA色选设备,主要包括:采用不同的滤波算法对图像进行去噪,并对处理结果应用边缘检测算子进行对比分析,最终选取并改进了中值滤波算法。采用图像分割算法得到二值化的稻米图像,再采用改进的形态学处理算法,去除图像中的小噪声及瑕疵点,平滑物体的轮廓。采用图像连通区域标记算法,对稻米图像中的各个籽粒进行有效识别。同时对稻米外观检测算法进行研究,通过轮廓跟踪获取稻米籽粒的边界轮廓,再求取最小外接矩形计算出籽粒的长短轴,从而得到粒型信息。通过设置合适的灰度阈值,实现对异色米的检测。分析不同图像处理算法的硬件实现的特点,将图像处理算法移植到嵌入式黑白色选系统上。采用VHDL硬件描述语言,在Xilinx公司的ISE13.4开发工具下编写改进的中值滤波优化算法、图像二值化算法、形态学处理算法和连通区域标记算法,并测试每个算法的处理效果和运行时间,最后实现稻米的粒型参数和异色米的检测。
[Abstract]:The color selection system based on machine vision is a kind of intelligent screening equipment, which involves optical imaging, image processing and electrical signal control. At present, the computer is used as the control core in the image signal processing in the color selection system, which has the disadvantages of large volume and low screening efficiency. Nowadays, the leapfrog development of embedded technology provides new ideas and methods for color separation system, and black and white color separation equipment has been widely used because of its advantages of simple structure and high performance to price ratio. Therefore, this paper studies the embedded black and white color selection system based on machine vision, with FPGA as the control core, which has the characteristics of good real-time and high reliability, and has a broad application prospect in the fields of industry and agriculture. In this paper, the main control circuit board and the output driving circuit board of FPGA are designed and fabricated by taking the black and white linear CCD camera as the sensing element and the XC3S500E chip of Xilinx Company as the processing core. The dual port RAM is built by IP technology of FPGA to realize the separation control of reading and storage of material information. Because the uneven illumination will bring deviation to the measurement result, the VHDL hardware description language is used to realize the correction function of the visual data, and the pixel gray value is compensated for nonlinearity. Reprogramming to achieve comparative judgment and delay output drive signal and other functions. The drive circuit controls the opening and closing of the valve to eliminate the defective products. Taking rice detection as the research object, using MATLAB simulation tools to study various image processing algorithms, and improve them to make them suitable for FPGA color sorting equipment, including: using different filtering algorithms for image denoising. Finally, the median filtering algorithm is selected and improved. The binary rice image is obtained by using image segmentation algorithm, and then the improved morphological processing algorithm is used to remove the small noise and defect points in the image and smooth the contour of the object. In this paper, the connected region labeling algorithm is used to identify the grains in the rice image. At the same time, the algorithm of rice appearance detection was studied. The boundary contour of rice grain was obtained by contour tracking, and the grain length axis was calculated by the smallest outer rectangle, and the grain shape information was obtained. By setting the appropriate gray threshold, the detection of different color rice can be realized. The characteristics of hardware implementation of different image processing algorithms are analyzed, and the image processing algorithm is transplanted to embedded black and white color selection system. Using VHDL hardware description language, the improved median filter optimization algorithm, image binarization algorithm, morphological processing algorithm and connected area marking algorithm are compiled under the ISE13.4 development tool of Xilinx Company. The processing effect and running time of each algorithm were tested.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】