视觉乐谱识别关键技术研究农业机械化工程专业论文.docxVIP

在的乐符镜像现象。该切分及识别方法有效解决了视觉乐谱图像谱线去除难题，提 在的乐符镜像现象。该切分及识别方法有效解决了视觉乐谱图像谱线去除难题，提 供了新的乐符定位及切分思路，提取了适应性强的乐符特征，提高了带有谱线乐符 的识别率。 关键词：视觉乐谱识别；非监督特征学习；多阈值二值化；拐点；区域增长算法； 方向梯度直方图 2 万方数据 Study Study the key technology of visual music image recognition Author：Yang Fang Supervisor：Teng Guifa Major：Engineering of Agricultural Mechanization Abstract Optical music recognition(OMR)technology converts digital music image into computer understanded code．This is of great significance to the music information precessing，to develop of rural music education and to improve farmer’S entertainment．It Can also provide technical support for妣music works selection for me agricultural production．The visual muisc image collected bv t11e mobile phone built—in camera iS more suitable for the working environment of agriculture，and the economic advantage iS obvious．Inflenced by focusing equipment shooting in the natural environment，visual music image is characteristed with complex background，uneven illumination，staff tilt，bending and fuzzy，which has seriously affected the music recognition effect．In view of the difficulty of visual staff music image recognition，music automatic position，image binarization，staff line detection and music character recognition are discussed in this paper to improve the system efhciency．The main contributions of Our work are listed as follow： (1) The unsupervised feature learning(UFL)method is introduced into the music score automatic position．Image block iS the discriminant unit in this framework， block feature is extracted by K—means clustering algorithm，support vector machine iS used as classification．music score location iS determined by block connected area searching．Compared with other features，the UFL feature has some advantages of the adaptability and complexity in visual music image process，the music area detection error rate iS reduced when block feature iS influenced by the Chianese characters． (2) A two—level area segmentation mathod based on伊ay difference is propo