摘要
棉花中的异性纤维给棉纺织企业带来了不小的损失，因此开发出高效率的异纤分拣机是很多纺织企业的迫切需求。现有的异纤分拣机虽然能够达到剔除部分异纤的效果，但由于图像处理运算都是在工控机上的CPU里实现，为了达到在线实时检测的需求，只能采用比较简单的识别算法，异纤识别效果不是很理想。本文通过研究基于CPU＋GPU异构系统的棉花异纤识别系统，让GPU通过并行运算来实现异纤的识别算法，从而大大提高了运算效率，有效的减少了算法的运行时间，为复杂的异纤识别算法在异纤识别系统的使用提供了条件，从而有效提高异纤杂质的剔除率。实验表明，基于CPU＋GPU异构系统对算法的时间提高了十几倍。
Foreign fibers in cotton bring huge loss for the textile enterprises.Therefore,the development of a high efficiency of foreign fiber sorter is the urgent needs of the many textile enterprises.Although the existing foreign fiber sorter can be able to reach excluding some part of foreign fibers,due to the image processing operations are carried out on CPU,in order to reach the demand for online real-time detection,only simple recognition algorithms are taked in,the result is not very satisfactory.In this paper, the cotton foreign fiber identification system based on the CPU＋GPU heterogeneous systems is researched,GPU parallel computing achieves the foreign fiber identification algorithm,thus the efficiency of operations is greatly improved,it effectively reduces the running time of the algorithm,offers the conditions in the use of a complex foreign fiber identification algorithm,and effectively improves the foreign fiber rejection rate.The experiment showed that heterogeneous system based on the CPU＋GPU can improve ten times in the time of algorithms.
出处
《现代科学仪器》
2014年第1期65-69,共5页
Modern Scientific Instruments
