基于线阵CCD图像亚像素级别边缘检测方法

中图分类号:TP23???????????????? 文献标识码:A

摘要:

研究线阵CCD采集一维图像亚像素级别边缘检测方法。该方法经过图像平滑滤波、基于梯度算子边缘粗定位、拟合区间搜索、最小二乘直线拟合边缘精确定位几个步骤，定位出线阵CCD图像边缘位置。图像采集卡(FPGA)接收图像处理卡(DSP)下发的控制指令，接收CCD图像数据，并且负责图像数据的平滑滤波和梯度值计算，粗定位出边缘位置，然后把边缘粗位置坐标，以及截取粗定位的边缘点附近段数据上传给图像处理卡，以供后者进行下一步拟合区间搜索和最小二乘拟合精确定位出边缘位置。通过图像采集卡的前期处理，可以大大减少其与图像处理卡之间的通讯数据量，FPGA+DSP这种分工处理模式发挥了各自处理芯片的优势，大大提高了运算处理效率。实验结果表明，该算法检测精度达到亚像素级别，3σ小于0.5像素，处理时间小于90μs。

Abstract:

This study was focused on a subpixel-level edge detection method for linear CCD images. The method detected the edge position of the array ccd image via several steps of image smooth filtering, edge coarse orientation based on gradient operator, fitting region searching and edge precise orientation based on least square straight-line fitting. The image acquisition card (FPGA) received the CCD image data after receiving the control command sent by the image processing card (DSP), and was responsible for the smooth filtering and gradient value calculation, positioning the rough edge coordinate, and then passed the coordinate and the data around it to the image processing card for the next fitting interval searching and the fine edge positioning. Through the early processing of the image acquisition card, the amount of communication data between it and the image processing card could be greatly reduced. This division of processing mode between FPGA and DSP took the advantages of their respective processing chips and greatly improved the processing efficiency. The experimental results proved the edge detecting precision had achieved sub-pixel level, 3σ was less than 0.5 pixel, the processing time was less than 90μs.