| 朱家微,江朝晖,邰其乐,岳旭东,洪石兰.基于热成像的动态多目标温度智能监测仪[J].安徽建筑大学学报,2021,29(): |
| 基于热成像的动态多目标温度智能监测仪 |
| Dynamic multi-target temperature intelligent monitor based on thermal imaging |
| 投稿时间：2021-02-22 修订日期：2021-04-25 |
| DOI：10.11921/j.issn.2095-8382.20210224001 |
| 中文关键词: 热红外成像 温度监测 MCU TensorFlow Lite 目标识别 |
| 英文关键词: thermal infrared imaging temperature monitoring MCU TensorFlow Lite object identification |
| 基金项目:安徽高校自然科学研究重大项目(KJ2019ZD20),智慧农业技术与装备安徽省重点实验室自主创新研究基金(APKLSATE2019X002)。 |
|
| 摘要点击次数: 1711 |
| 全文下载次数: 537 |
| 中文摘要: |
| 为了满足小型室内场所人员体温自动监测的需求,设计一款低成本的动态多目标温度智能监测仪。采用高分辨率微型热成像模块Lepton 3.5获取热红外图像,在STM32H743上搭载TensorFlow Lite深度学习框架,检测出热红外图像中的人体并标记其温度,重点基于“滑动窗口”技术设计新的人体识别算法,解决了由于MCU算力不足导致的视频卡顿问题。实验结果表明,目标识别的准确率高达99.7%,监测多个动态目标温度时视频流畅。该温度监测仪体积小,成本低,检测目标多,适用于家庭、办公室和实验室等场所。 |
| 英文摘要: |
| In order to automatically monitor the temperature of people in small indoor places, a low cost dynamic multi-target intelligent temperature monitor was designed. Lepton 3.5, a high-resolution miniature thermal imaging module, was used to acquire thermal infrared images. TensorFlow Lite deep learning framework was mounted on STM32H743 to detect the human body in the thermal infrared images and mark its temperature. A new human body recognition algorithm was designed based on the sliding window, which was used to solve the problem of video lull caused by insufficient MCU computing power. The experimental results show that the accuracy of target recognition was as high as 99.7%, and the video was smooth when monitoring multiple dynamic target’s temperatures. The temperature monitor has the advantages of small size, low cost and many detecting targets. It is suitable for home, office , laboratory and so on. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
