| 引用本文: | 王梓,毕道明,周颉鑫,等.结构主导目标检测中的纹理随机化.[J].国防科技大学学报,2021,43(4):24-30.[点击复制] |
| WANG Zi,BI Daoming,ZHOU Jiexin,et al.Texture randomization in structure-dominant object detection[J].Journal of National University of Defense Technology,2021,43(4):24-30[点击复制] |
|
| |
|
|
| 本文已被：浏览 6910次 下载 5115次 |
| 结构主导目标检测中的纹理随机化 |
| 王梓1,2,毕道明1,3,周颉鑫1,2,孙晓亮1,2,于起峰1,2 |
|
(1. 国防科技大学 空天科学学院, 湖南 长沙 410073;2. 国防科技大学 图像测量与视觉导航湖南省重点实验室, 湖南 长沙 410073;3. 沈阳飞机设计研究所, 辽宁 沈阳 110035)
|
| 摘要: |
| 已有基于卷积神经网络的目标检测算法倾向于提取目标纹理特征,而非结构特征；因此,已有方法不能实现变纹理目标的可靠检测。针对此问题,提出基于纹理随机化的结构主导目标检测方法,采用仿真纹理随机化方法减弱网络模型对纹理特征的拟合,实现基于结构特征的变纹理目标可靠检测。利用目标的三维模型,借助Blender渲染引擎,完成纹理随机化仿真训练数据集的生成。仿真及真实图像实验测试结果表明:该方法能够实现基于目标结构特征的变纹理目标可靠检测。 |
| 关键词: 纹理随机化 仿真数据 目标检测 三维模型 |
| DOI：10.11887/j.cn.202104004 |
| 投稿日期：2019-12-20 |
| 基金项目:湖南省自然科学基金资助项目(2019JJ50732) |
|
| Texture randomization in structure-dominant object detection |
| WANG Zi1,2, BI Daoming1,3, ZHOU Jiexin1,2, SUN Xiaoliang1,2, YU Qifeng1,2 |
|
(1. College of Aerospace Science and Engineering, National University of Defense Technology, Changsha 410073, China;2. Hunan Key Laboratory of Videometrics and Vision Navigation, National University of Defense Technology, Changsha 410073, China;3. Shenyang Aircraft Design & Research Institute, Shenyang 110035, China)
|
| Abstract: |
| Region-based detectors with convolutional neural networks tend to learn textural rather structural feature and thus face substantial difficulties in detecting objects with various textures. To tackle this problem, the texture randomization to augment the synthetic training image dataset was employed and a novel method for structure-aware object detection was proposed. The texture-randomized simulation data were generated by rendering 3D model with varied textures using Blender. Experiments on synthetic and real images indicate that the proposed method is capable of robustly detecting texture-varied objects based on structural information. |
| Keywords: texture randomization simulation data object detection 3D model |
|
|
|
|
|
