ID 原文 译文
25145 乐谱图像的自动分割、倾斜校正是乐谱识别过程中的关键技术，各种计算机光学乐谱识别技术在乐谱图像的数字化中有着广泛的应用，但对于乐谱中简谱的识别一直鲜有研究。 Automatic segmentation and tilt correction of music images are key techniques in music recognition. Various computer optical music score recognition technologies have been widely used in the digitization of music score images, but there has been little research on the recognition of numbered musical notation．
25146 本文针对人工拍摄条件下光照不理想的简谱图像，提出一种基于 PCNN(脉冲耦合神经网络)和 DNN(深度神经网络)相结合的分块简谱图像自动分割算法，该方法根据简谱图像灰度分布特征对图像进行自适应分块处理，依据每个分块的灰度特征与 PCNN 最佳迭代次数之间的关系构造合适的 DNN 神经网络，从而实现了最优分割图的自适应选取; In this paper, an automatic image segmentation algorithm based on PCNN (Pulse Coupled Neural Networks) and DNN (Deep Neural Networks) is proposed to solve the problem of a variety of lighting conditions. The image is processed by adaptive block processing according to the gray scale distribution of the spectral image and analyze the relationship between the gray scale characteristics of each block and the optimal PCNN iteration time, construct an appropriate DNN neural network to realize the adaptive selection of optimal segmentation graph.
25147 进一步利用最优分割图像中音符小节线的水平投影，提出一种双尺度下降法实现了简谱图像的倾斜校正; Further using the horizontal projection of bar lines, we propose a dual-scale descent method to realize the skew correction of numbered musical notation image．
25148 提出去边垂直投影法和连通域距离判断法实现了简谱图像中音符及歌词的提取。 We propose an edgeless vertical projection method and connected domain distance judgment method to extract the note and lyrics from numbered musical notation image．
25149 实验仿真结果表明:本文算法对复杂光照条件下的简谱图像处理都具有较好的鲁棒性，同时表现出更高的效率。 Simulation experiments show that the proposed algorithm exibits better robustness for numbered musical notation image under complex illumination conditions and faster speed．
25150 针对单阈值-非极大值抑制算法中出现的目标漏检和重复检测问题，本文提出了一种使用全局交并比指标 GIoU(Generalized Intersection over Union)衡量目标相似度的双阈值非极大值抑制算法 GDT-NMS(Generalized DualThreshold NMS，GDT-NMS)。 Aiming at the problem of missed detection and repeated detection in the single-threshold-non-maximum suppression algorithm, this paper proposes a dual-threshold Non-Maximum Suppression algorithm using GIoU (Generalized Intersection over Union).
25151 使用双阈值改进 NMS 算法和 soft-NMS 算法，抑制多余的检测框; Using dual thresholds to improve the NMS algorithm and the soft-NMS algorithm, suppressing redundant detection boxes．
25152 在此基础上，使用 GIoU替换传统的 IoU 计算目标间的相似度，使目标的定位更加准确; Based on the above, using GIoU instead of IoU to calculate the similarity between objects，so that the positioning of the object is more accurate;
25153 进一步，使用非线性函数赋予检测框不同比例的权值惩罚，使检测框的得分随距离呈非线性变化，目标区分度更高。 the non-linear function is used to give different weights to the proposal boxes, which makes the proposal boxes' scores change non-linearly with distance, and the target discrimination is higher, which is more conducive to suppressing the proposal boxes．
25154 改进算法在 PASCAL VOC 和 MSCOCO 上的检测精度分别为 74. 8% 和 25. 9%。 The detection accuracy of the improved algorithm on PASCAL VOC and MSCOCO is 74.8% and 25.9% , respectively.