Accelerated hardware target labeling based on run-length accumulation
-
摘要: 针对目标标记算法的长耗时、大延迟严重影响目标跟踪的效果和实时性问题,提出了快速的基于行程累加的目标标记硬件加速算法.通过对像素进行行程编码,按照定义的累加规则对相邻行的行程进行累加运算,避免了标号冲突的发生,无需建立等价表,实现了一遍扫描完成连通域标记和连通域特征的计算.实验结果表明提出的算法资源占用较少,输出延迟小,与软件算法相比具有很高的加速比,有利于后续目标跟踪的高速实时处理.Abstract: To solve the problem that long time consuming and delay of target labeling algorithm seriously affected the effect and real-time performance of target tracking, fast accelerated hardware target labeling algorithm based on run-length accumulation was proposed. The pixels of image were converted to runs, and then the two runs of adjacent lines were accomplished accumulation calculation according to the defined accumulation rulers. The labeling conflicts were avoided. The algorithm could complete connected components labeling and calculate connected components features in a single pass without establishing equivalence table. Experiments show that the algorithm uses less memory resources with smaller result delay, and has higher speed ratio compared with software algorithms. So it can benefit to the fast and real-time processing of later target tracking.
-
Key words:
- parallel computation /
- target labeling /
- accelerated hardware /
- run-length accumulation
-
[1] Khanna V,Gupta P,Hwang C J.Finding connected components in digital images [C]//Henry Selvaraj.International Conference on Information Technology.Las Vegas:IEEE,2005:652-656 [2] Suzuki K,Horiba I,Sugie N.Linear-time connected-component labeling based on sequential local operations[J].Computer Vision and Image Understanding,2003,89(1):1-23 [3] Suzuki K,Horiba I,Sugie N.Fast connected component labeling based on sequential local operations in the course of forward raster scan followed by backward raster scan [C]//Bob Werner.15th International Conference on Pattern Recognition.Barcelona:IEEE,2000:434-437 [4] Ito Y,Nakano K.Optimized component labeling algorithm for using in medium sized FPGAs [C]//John Hine.Ninth International Conference on Parallel and Distributed Computing.Dunedin:IEEE,2008:171-176 [5] 周平,王肖玉,刘越,等.并行流水线实现的多红外标志点实时标记与识别新方法[J].仪器仪表学报,2009,30(6):1240-1244 Zhou Ping,Wang Xiaoyu,Liu Yue,et al.New real-time labeling and recognition method for multiple infrared markers based on parallel pipeline[J].Chinese Journal of Scientific instrument,2009,30(6):1240-1244(in Chinese) [6] Ma Ni,Bailey D G,Johnston C T.Optimised single pass connected components analysis [C]//Tarek Ghazawi.International Conference on Field-Programmable Technology.Taipei:IEEE,2008:185-192 [7] He Lifeng,Chao Yuyan,Kenji Suzuki.A linear-time two-scan labeling algorithm [C]//Jeff Rodriguez.International Conference on Image Processing.San Antonio:IEEE,2007:241-244 [8] He Lifeng,Chao Yuyan,Kenji Suzuki,et al.Fast connected-components labeling[J].Pattern Recognition,2009,42(9):1977-1987 [9] Roy Pradipta,Biswas P K.VLSI implementation of fast connected component labeling using finite state machine based cell network [C]//Jayanta Mukherjee.Computer Vision Graphics image Processing.Bhubaneswar:IEEE,2008:238-243 [10] Johnston Christopher T,Bailey Donald G.FPGA implementation of a single pass connected components algorithm [C]//Amine Bermark.IEEE International Symposium on Electronic Design.DELTA:IEEE,2008:228-231 [11] 孔斌.快速连通域分析算法及其实现[J].模式识别与人工智能,2003,16(1):110-115 Kong Bin.A fast connected component analysis algorithm and its implementation[J].PR&AI,2003,16(1):110-115(in Chinese) [12] Appiah K,Hunter A,Dickinson P,et al.Accelerated hardware video object segmentation:From foreground detection to connected components labeling[J].Computer Vision and Image Understanding,2010,114(11):1282-1291
点击查看大图
计量
- 文章访问数: 1655
- HTML全文浏览量: 19
- PDF下载量: 691
- 被引次数: 0