FPGA的高速并行性、流水线方式处理数据的特点，适用快速图像处理。而在图像处理应用中，图像的关键在于边缘特征所包含的信息。边缘检测在目标跟踪、目标识别、深度学习下的监督识别、机器视觉检测等领域至关重要。在低照度环境下，图像的边缘信息不明显，精确地检测出低照度环境下的边缘非常关键。

针对低照度环境下引起的图像质量退化导致边缘检测精度低的问题，提出一种基于梯度差分自适应边缘检测方法，选取Canny算子作为边缘检测算子并做出改进。采用双边滤波器来替代传统的高斯滤波器，增加45°和135°两方向的梯度计算模板来突出边缘；针对Canny阈值选择不具有自适应性的特点，基于梯度差分对Otsu算法进行改进得到自适应阈值进行分割，分割后采用递归边界跟踪的方法连接边缘。在此基础上基于FPGA设计实现了一个实时边缘检测系统，通过配置OV5640摄像头进行图像数据采集，为了提高低照度环境下的图像质量，通过调节白平衡和Gamma校正的参数来调节图像质量，图像增强中加入四方向拉普拉斯锐化进突出边缘信息，并将改进Canny算法进行硬件设计实现，最后通过VGA显示图像和检测的边缘信息。本文在系统中使用SDRAM DDR3 芯片缓存图像数据，提高系统的鲁棒性和数据吞吐量。

通过SNR（信噪比）和C/A、C/B的比值两种指标评判边缘检测效果，测试结果表明相比于传统Canny算法，改进Canny在低照度环境下的图像边缘检测效果较好，边缘清晰且连续，自适应阈值判断时间仅为原来的十分之一，更适合实时运算。整个实时边缘系统在ATRIX-7开发板完成搭建，实验平台为Vivado 2019.1，通过与Sobel和传统Canny算法的实时处理来评判实时边缘检测效果，测试结果表明，本文系统检测的边缘噪声少，弱边缘提取效果好，边缘完整且连续，本文的实时处理速度可以达到60帧/秒，实时性较好。

FPGA is suitable for fast image processing because of its high-speed parallelism and pipeline data processing. In the application of image processing, the key of image is the information contained in the edge features. Edge detection is very important in target tracking, target recognition, supervised recognition under deep learning and machine vision detection. In the low illumination environment, the edge information of the image is not obvious, so it is very important to accurately detect the edge in the low illumination environment.

Aiming at the problem of low accuracy of edge detection caused by image quality degradation in low illumination environment, an adaptive edge detection method based on gradient difference is proposed. Canny operator is selected as the edge detection operator and improved. A bilateral filter is used to replace the traditional Gaussian filter, and gradient calculation templates in 45 ° and 135 ° directions are added to highlight the edge. In view of the fact that canny threshold selection is not adaptive, Otsu algorithm is improved based on gradient difference to obtain adaptive threshold for segmentation. Finally, recursive boundary tracking method is used to connect the edges. On this basis, a real-time edge detection system is designed and implemented based on FPGA. The OV5640 camera is configured to collect image data. In order to improve the image quality in low illumination environment, the parameters of white balance and gamma correction are adjusted to adjust the image quality. Four direction Laplacian sharpening is added to the image enhancement to highlight the edge information, and the improved Canny algorithm is hard coded Finally, VGA is used to display the image and detect the edge information. In this paper, SDRAM DDR3 chip is used to cache image data in the system to improve the robustness and data throughput of the system.

SNR (signal-to-noise ratio) and the ratio of C/A and C/B are used to evaluate the edge detection effect. The test results show that compared with the traditional Canny algorithm, the improved Canny algorithm has better edge detection effect in low illumination environment, with clear and continuous edges. The adaptive threshold judgment time is only one tenth of the original, which is more suitable for real-time operation. The whole real-time edge system is built on the atrix-7 development board, and the experimental platform is vivado In January, 2019, the real-time edge detection effect is evaluated by real-time processing with Sobel and traditional Canny algorithm. The test results show that the edge noise detected by this system is less, the weak edge extraction effect is good, and the edge is complete and continuous. The real-time processing speed of this paper can reach 60 frames per second, and the real-time performance is good.

[1]Samsir et al. Edge Detection to Make Drawing Sketch using Laplacian Operator and Gabor Wavelet for Learning Devices[J]. Journal of Physics: Conference Series, 2021, 1764(1) : 012070-.

[2]Akshi Kumar and Sahil Raheja. Edge Detection using Guided Image Filtering and Enhanced Ant Colony Optimization[J]. Procedia Computer Science, 2020, 173 : 8-17.

[3]Cao Yiqin and Wu Dan and Duan Yeyu. A new image edge detection algorithm based on improved Canny[J]. Journal of Computational Methods in Sciences and Engineering, 2020, 20(2) : 629-642.

[4]Li Ting et al. Enhanced optical edge detection based on a Pancharatnam-Berry flat lens with a large focal length.[J]. Optics letters, 2020, 45(13) : 3681-3684.

[5]王闯, 贺莹, 余云,等. 一种基于FPGA的全帧率图像处理方法[J]. 航空计算技术, 2013（06）:112-117.

[6]曾建华,黄时杰.典型图像边缘检测算子的比较与分析[J].河北师范大学学报（自然科学版）,2020,44（04）:295-301.

[7]Shi Guiming,Suo Jidong. Remote Sensing Image Edge-Detection Based on Improved Canny Operator[A]. IEEE.Proceedings of 2016 8th IEEE International Conference on Communication Software and Networks (ICCSN 2016)[C].IEEE:IEEE BEIJING SECTION(跨国电气电子工程师学会北京分会),2016:5.

[8]李庆忠,刘洋.基于改进Canny算子的图像弱边缘检测算法[J].计算机应用研究,2020,37（S1）:361-363.

[9]赵娅琳, 陶青川, 卢倩雯等. 基于梯度二阶导数的Canny阈值自适应选取算法[J]. 科学技术与工程, 2018, 018（014）:179-185.

[10]范艳军,吴晓红,罗代升.一种改进的基于小波变换的图像边缘检测算法[J].四川大学学报（自然科学版）,2012,49（06）:1264-1268.

[11]田国富,宋江波.基于多尺度多结构元的数学形态学边缘检测[J].微型机与应用,2016,35（07）:41-43.

[12]李成,彭良玉.SUSAN边缘检测算法的设计与研究[J].微型机与应用,2014,33（05）:37-39.

[13]李颖杰,杨华.基于改进SUSAN算法的红外图像边缘增强方法[J].光电技术应用,2015,24（04）:64-66.

[14]曾雪琴. 基于改进遗传神经网络的深度图像边缘检测研究[D].哈尔滨理工大学,2012.

[15]冯会真,夏哲雷,林志一.基于神经网络的图像边缘检测方法[J].中国计量学院学报,2012（04）:289-291.

[16]李忠虎,张琳,闫俊红.管道腐蚀视觉测量图像边缘检测算法研究[J].电子测量与仪器学报,2017,31（11）:1788-1795.

[17]曾希君,于博.基于改进BP神经网络图像边缘检测的研究[J].微电子学与计算机,2012,26（08）:215-218.

[18]闫河,赵其峰,谢敏,李晓玲.自适应Canny算子PCBA元器件边缘检测及修复[J].光学学报,2021,41(05):97-104.

[19]Talal Bonny and Safaa Henno. Image Edge Detectors under Different Noise Levels with FPGA Implementations[J]. Journal of Circuits, Systems and Computers, 2018, 27(13)

[20]Jian Guo Yang and Bei Zhi Li and Hua Jiang Chen. Adaptive Edge Detection Method for Image Polluted Using Canny Operator and Otsu Threshold Selection[J]. Advanced Materials Research, 2013, 1359 : 797-804.

[21]Ratsakou Almpion et al. Shape reconstruction of delamination defects using thermographic infrared signals based on an enhanced Canny approach[J]. Infrared Physics & Technology, 2020, 111 : 103527-.

[22]Yue S , Yuan M , Lu T , et al. Rotation Invariant Angle-Density based Features for an Ice Image Classification System[J]. Expert Systems with Applications, 2020, 162:113744.

[23]M F V Ruslau et al. Edge detection of digital image with different edge types[J]. Journal of Physics: Conference Series, 2020, 1569(4) : 042069-.

[24]Nsd A , Pkd B . Edge Detection Technique using Binary Particle Swarm Optimization - ScienceDirect[J]. Procedia Computer Science, 2020, 167:1421-1436.

[25]李长有,陈国玺,丁云晋.改进Canny算子的边缘检测算法[J].小型微型计算机系统,2020,41（08）:1758-1762.

[26]李正,张海.一种基于梯度方向一致性的边缘检测新算法[J].电子与信息学报,2016,38（07）:1674-1681.

[27]何银水,李岱泽,赵梓宇,钱韦旭.基于亮度突变性与密度特征检测的厚板T形接头焊缝轮廓识别[J/OL].上海交通大学学报:1-7[2021-04-

[28]段红燕,邵豪,张淑珍等.一种基于Canny算子的图像边缘检测改进算法[J].上海交通大学学报,2016,50（12）:1861-1865.

[29]张桂梅,孙晓旭,陈彬彬等.结合分数阶微分和Canny算子的边缘检测[J].中国图象图形学报,2016,21（08）:1028-1038.

[30]宋人杰,刘超,王保军.一种自适应的Canny边缘检测算法[J].南京邮电大学学报（自然科学版）,2018,38（03）:72-76.

[31]张帆, 彭中伟, 蒙水金. 基于自适应阈值的改进Canny边缘检测方法[J]. 计算机应用, 2012（08）:2296-2298.

[32]李健,刘孔宇,任宪盛,熊琦,窦雪峰.基于自适应阈值的Canny算法在MRI边缘检测中的应用[J].吉林大学学报（工学版）,2021,51（02）:712-719.

[33]薛丽霞,李涛,王佐成.一种自适应的Canny边缘检测算法[J].计算机应用研究,2014,27（09）:3588-3590.

[34]Singh S , Saini A K , Saini R , et al. A novel real-time resource efficient implementation of Sobel operator-based edge detection on FPGA[J]. International Journal of Electronics, 2014, 101(10-12):1705-1715.

[35]Xu Q , Varadarajan S , Chakrabarti C , et al. A Distributed Canny Edge Detector: Algorithm and FPGA Implementation[J]. IEEE Transactions on Image Processing, 2014, 23(7):2944-2960.

[36]Possa P R , Mahmoudi S A , Harb N , et al. A Multi-Resolution FPGA-Based Architecture for Real-Time Edge and Corner Detection[J]. IEEE Transactions on Computers, 2013.

[37]Nausheen, Nazma, Seal, et al. A FPGA based implementation of Sobel edge detection[J]. Microprocessors and microsystems, 2018, 56(Feb.):84-91.

[38]Engineering; Study Findings on Engineering Are Outlined in Reports from Kuwait University (Implementation of an Edge Detection Algorithm Using Fpga Reconfigurable Hardware)[J]. Mathematics Week, 2020, : 991-.

[39]Saeid Taslimi et al. Adaptive Edge Detection Technique Implemented on FPGA[J]. Iranian Journal of Science and Technology, Transactions of Electrical Engineering, 2020, : 1-12.

[40]Dong E , Zhao Y , Yu X , et al. An Improved NMS-Based Adaptive Edge Detection Method and Its FPGA Implementation[J]. Journal of Sensors,2016,(2015-12-22), 2015, 2016:1-8.

[41]胡晓莉,丁丽,江杰.改进Canny算法的实时图像边缘检测[J].内蒙古科技大学学报,2015,34（03）:262-266.

[42]陈鑫元,李筠,杨海马,宋佳,邵璐先.自适应阈值图像二值化及形态学处理的FPGA实现[J].电子测量技术,2016,39（07）:67-71.

[43]杜建宝,董恩增,张祖锋.基于形态学的边缘检测系统与FPGA实现[J].天津理工大学学报,2018,34（04）:26-29+39.

[44]史伟忠,曹卫卫,范彦铭,董佳筠,陈纾,肖昊.基于FPGA的深空图像实时边缘检测算法与实现[J].电子科技,2020,33（05）:45-49.

[45]张灏,王素珍,郑宇,王伟,李潍志,王鹏,任贵珊,马家麟.一种组合GAUSS-filter、SOBEL、NMS、OTSU 4种算法的图像边缘检测的FPGA实现[J].液晶与显示,2020,35（03）:250-261.

[46]Nasor Mohamed and Obaid Walid. Segmentation of osteosarcoma in MRI images by K‐means clustering, Chan‐Vese segmentation, and iterative Gaussian filtering[J]. IET Image Processing, 2020, 15(6) : 1310-1318.

[47]Yu Dongling et al. Multi-scale decomposition enhancement algorithm for surface defect images of Si;N;ceramic bearing balls based on stationary wavelet transform[J]. Advances in Applied Ceramics, 2021, 120(1) : 47-57.

[48]Sivakumar V , Janakiraman N . A novel method for segmenting brain tumor using modified watershed algorithm in MRI image with FPGA[J]. Biosystems, 2020, 198(2):104226.

[49]Saeid Taslimi et al. Adaptive Edge Detection Technique Implemented on FPGA[J]. Iranian Journal of Science and Technology, Transactions of Electrical Engineering, 2020, : 1-12.

[50]Cao J , Chen L , Min W , et al. Implementing a Parallel Image Edge Detection Algorithm Based on the Otsu-Canny Operator on the Hadoop Platform[J]. Computational Intelligence & Neuroscience, 2018, 2018:1-12.

[51]薛丽霞,李涛,王佐成.一种自适应的Canny边缘检测算法[J].计算机应用研究,2015,27(09):3588-3590.

[52]陆涛.基于统计特征分类耦合自适应Gamma校正的图像增强算法[J].电子测量与仪器学报,2020,34(06):154-162.

[53]魏亚,孔维康,万成,左勇志,鲁巧稚.比色法检测受火后混凝土损伤程度[J].吉林大学学报(工学版),2021,51(01):233-244.