| 胡春海,姜 楠,刘 斌.基于旷场实验的社交行为分类学习方法[J].高技术通讯(中文),2023,33(5):489~496 |
| 基于旷场实验的社交行为分类学习方法 |
| Classification learning method of social behavior based on open field experiment |
| |
| DOI：10. 3772/ j. issn. 1002-0470. 2023. 05. 005 |
| 中文关键词: 旷场实验; 社交行为; 关键点检测; 虚拟柔态模型 |
| 英文关键词: open field test, social behavior, key points detection, virtual flexible model |
| 基金项目: |
| 作者 | 单位 | | 胡春海 | (燕山大学电气工程学院 秦皇岛066000) | | 姜 楠 | (燕山大学电气工程学院 秦皇岛066000) | | 刘 斌 | (燕山大学电气工程学院 秦皇岛066000) |
|
| 摘要点击次数: 1063 |
| 全文下载次数: 803 |
| 中文摘要: |
| 旷场实验是动物行为学与药物实验分析中常用的实验方法。为了获取药物对小鼠情感变化的影响程度,通常需要多只小鼠来进行实验。在记录多只小鼠行为状态的过程中,不可避免地会产生遮挡问题,导致对小鼠行为差异的观测精确度降低,从而影响对药物疗效的判断。为了解决此问题,本文以真实状态下拍摄的数据集作为研究对象,提出虚拟柔态模型来对遮挡部分信息进行预测,对所分类的特征进行数据增强,进而提高了社交行为分类的准确率。在测试中,该方法对于3 种行为分类的准确率分别为92. 8%、99. 7%、99. 0%。另外,本文提出的虚拟柔态模型是基于单相机拍摄,有效减少了拍摄时的人为误差以及成本耗费,其社交行为分类准确率与三维空间拍摄的分类准确率相似,但在硬件设施上以及数据集处理方面优于现有的分类方法。 |
| 英文摘要: |
| Open field experiment is a common experimental method in animal behavior and drug experimental analysis.Experiments with multiple mice are usually carried out to determine the extent to which a drug affects emotional changes between mice. In the process of recording the behavior state of multiple mice, it is inevitable to encounter the problem of occlusion, which leads to the low accuracy of the observation of behavioral differences of mice and thus affects the judgment of drugs. In order to solve this problem, this paper takes the data set taken in the real state as the research object, and proposes a virtual flexible model to predict the occluded information and enhance the data of the classified features, thus improving the accuracy of social behavior classification. In the test, the classification accuracy of the three behaviors is 92. 8%, 99. 7% and 99. 0% respectively. In addition, the virtual flexible model proposed in this paper is based on single camera shooting, which reduces the human error and cost during shooting. The accuracy of the social behavior classification of mice is similar to that of the behavior classification after three-dimensional shooting. But the method proposed in this paper is superior to the existing classification methods in hardware facilities and data set processing. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
