一种改进的用于图像目标提取的多示例学习算法
文献类型：会议论文
| 作者 | 王孟月; 张常麟; 宋彦 |
| 出版日期 | 2010 |
| 会议名称 | 2010 Chinese Conference on Pattern Recognition (CCPR 2010) |
| 会议日期 | October 21-23, 2010 |
| 会议地点 | Chongqing, China |
| 关键词 | 多示例学习 视觉关键词 分割 目标提取 |
| 页码 | 5 pp. |
| 其他题名 | An Improved Multiple Instance Learning Algorithm for Object Extraction |
| 中文摘要 | 本文在MILES算法的基础上，提出了一种利用视觉关键词辞典为特征空间的多示例学习算法，并在示例判定的过程中结合分割实现了目标检测与提取。该方法采用“Bag of Words”模型，将图像作为多示例包，表示该图像的若干视觉关键词作为包中示例，并把视觉关键词辞典作为特征空间，通过对包中示例个数统计将其映射到特征空间中，随后采用1-normSVM来挑选重要特征同时对图像进行分类；对判定为正的图像进行示例判定，以判定为正的示例作为相应的目标“种子”点，然后进一步结合图像分割结果，实现目标提取。在Caltech101等标准图像集上的实验结果证明了本文方法的有效性。 |
| 英文摘要 | Based on MILES algorithm, we propose a novel multiple instance learning approach which regards visual word dictionary as feature space, and combines segmentation for object detection and extraction in the process of instance classification. This approach uses "Bag of Words" model. The whole image is considered as a multiple instance bag. The visual words that represent the image are regarded as the instances in the bag. The approach maps each bag into a feature space defined by visual vocabulary via the histogram over visual words. Next, 1-norm SVM is applied to select important features as well as classify images simultaneously. Then we will classify instances coming from the bags classified as positive, and take the positive instances for object "seed" points. After that segmentation is combined to realize object extraction. Experiments on Caltech101 dataset show that this approach achieves high efficiency. |
| 收录类别 | EI |
| 产权排序 | 2 |
| 会议录 | 2010 Proceedings of Chinese Conference on Pattern Recognition (CCPR 2010)
 |
| 会议录出版者 | IEEE |
| 会议录出版地 | Piscataway, NJ, USA |
| 语种 | 中文 |
| ISBN号 | 978-1-4244-7209-3 |
| 源URL | [http://ir.sia.cn/handle/173321/8960]  |
| 专题 | 沈阳自动化研究所_机器人学研究室 |
| 推荐引用方式 GB/T 7714 | 王孟月,张常麟,宋彦. 一种改进的用于图像目标提取的多示例学习算法[C]. 见:2010 Chinese Conference on Pattern Recognition (CCPR 2010). Chongqing, China. October 21-23, 2010. |
入库方式： OAI收割
来源：沈阳自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。