| 基于似物性的快速视觉目标识别算法 |
| |
| 引用本文： | 刘涛,吴泽民,姜青竹,曾明勇,彭韬频.基于似物性的快速视觉目标识别算法[J].计算机科学,2016,43(7):73-76, 94. |
| |
| 作者姓名： | 刘涛 吴泽民 姜青竹 曾明勇 彭韬频 |
| |
| 作者单位： | 中国人民解放军理工大学通信工程学院 南京210007,中国人民解放军理工大学通信工程学院 南京210007,中国人民解放军理工大学通信工程学院 南京210007,中国人民解放军理工大学通信工程学院 南京210007,中国人民解放军理工大学通信工程学院 南京210007 |
| |
| 基金项目： | 本文受航空科学基金(0125186005),国家自然科学青年基金(61501509)资助 |
| |
| 摘 要： | 针对视觉目标识别算法实时性较差的问题,基于似物性提出一种面向视觉目标识别的可变部件模型改进算法。该算法首先对图像进行二进制归一化的似物性检测,并利用检测结果形成视觉目标候选框；然后使用目标识别算法对候选区域进行似然判决,比滑动窗口法缩短了搜索时间；最后通过一个快速扩大-缩小算法对检测目标进行尺度修正,提高目标框的准确度。在PASCAL 图像库上的识别结果表明:该识别方法在准确率上优于当前主流的检测模型,计算耗时较级联DPM算法减少约50%。
|
| 关 键 词： | 计算机视觉 目标识别 似物性 可变部件模型 二进制梯度归一化 |
| 收稿时间： | 2015/9/27 0:00:00 |
| 修稿时间： | 2015/11/23 0:00:00 |
Fast Object Recognition Method Based on Objectness |
| |
| LIU Tao,WU Ze-min,JIANG Qing-zhu,ZENG Ming-yong and PENG Tao-pin.Fast Object Recognition Method Based on Objectness[J].Computer Science,2016,43(7):73-76, 94. |
| |
| Authors: | LIU Tao WU Ze-min JIANG Qing-zhu ZENG Ming-yong and PENG Tao-pin |
| |
| Affiliation: | College of Communications Engineering,PLA University of Science and Technology,Nanjing 210007,China,College of Communications Engineering,PLA University of Science and Technology,Nanjing 210007,China,College of Communications Engineering,PLA University of Science and Technology,Nanjing 210007,China,College of Communications Engineering,PLA University of Science and Technology,Nanjing 210007,China and College of Communications Engineering,PLA University of Science and Technology,Nanjing 210007,China |
| |
| Abstract: | In order to solve the poor real-time performance in object recognition,a fast object recognition method was proposed based on objectness.First,binarized normed gradients algorithm is used for a test image to get objectness eva-luations.Then,calculating with the objectness evaluations,a candidate bounding box is extracted.And next,deformable part model (DPM) algorithm is used to predict the object with image regions of the box,which can save time on gliding windows searching.Finally,a quickly expansion-shrinking procedure is used to modify the output boxes of DPM,improving accuracy.Experimental results on the challenging PASCAL VOC 2007 database demonstrate that the proposed method outperforms the state-of-art detection models in accuracy and almost twice faster than cascade DPM in instantaneity. |
| |
| Keywords: | Computer vision Object recognition Objectness Deformable part model (DPM) Binarized normed gradients (BING) |
|
|
点击此处可从《计算机科学》下载全文 |
|
