一个基于深度学习的动态可变形适应框架,用于定位动态火焰的根区域
Hongkang Tao1, Guhong Wang1, Jiansheng Liu1,2
1School of Advanced Manufacturing, Nanchang University, Nanchang, China.
PloS one
|April 17, 2024
概括
这项研究引入了一种新的动态可变形适应框架 (DDAF),用于精确的动态火焰检测. DDAF方法增强了火焰根的定位,改善了复杂环境中的火灾控制.
科学领域:
- 计算机视觉 计算机视觉
- 人工智能的人工智能
- 消防安全工程 消防安全工程
背景情况:
- 传统的光学火焰探测器 (OFD) 受到环境干扰,导致检测错误.
- 现有的深度学习模型在火焰识别方面表现出色,但在动态火焰源定位方面扎.
- 复杂的环境对准确和强大的火焰检测系统构成挑战.
研究的目的:
- 提出一个新的动态可变形适应框架 (DDAF),用于准确的动态火焰检测和定位.
- 解决现有模型在捕获动态火焰特征和根区域方面的局限性.
- 在复杂的工业和安全场景中提高火焰检测的精度和稳定性.
主要方法:
- 引入了可变形卷积网络v2 (DCNv2) 用于动态火焰的自适应特征提取.
- 综合上下文增强模块 (CAM) 和动态头 (DH) 用于多方面火焰特征分析.
- 采用基于幅度的层适应修剪 (LAMP) 来优化模型检测速度.
- 开发了具有粗和细粒度定位的诱导建模 (IM),用于精确的火焰根划分.
- 利用基于时间一致性的检测 (TCD) 来利用时间信息来提高稳定性.
主要成果:
- 与经典的深度学习方法相比,DDAF方法显示AP0.5提高了4.4%.
- 实现了模型参数 (25.3%) 和FLOP (25.9%) 的显著降低,提高了效率.
- 成功地以高精度动态定位了火焰根区域.
- 在定制火焰数据集上的实验结果验证了框架的有效性.
结论:
- 拟议的DDAF框架为动态火焰检测提供了强大而高效的解决方案.
- 该方法显著提高了复杂环境中的火焰根定位的准确性和速度.
- 这项研究将应用范围扩展到工业安全和燃烧过程控制等关键领域.
相关概念视频
Flame Photometry: Overview
573
Flame photometry, also known as flame emission spectrometry, is a technique used for the qualitative and quantitative analysis of elements present in a sample using a flame as the source of excitation energy. The concept of flame photometry was realized in the early 1860s by Kirchhoff and Bunsen, who discovered that specific elements emit characteristic radiation when excited in flames. The first instrument developed for this purpose was used to measure sodium (Na) in plant ash using a Bunsen...
573
Flame Photometry: Lab
244
In a flame photometer, when a solution like potassium chloride is aspirated into the flame, the solvent evaporates, leaving behind dehydrated salt. This salt dissociates into free gaseous atoms in their ground state. Some of these atoms absorb energy from the flame, leading to their excitation. The excited atoms return to the ground state, emitting photons at characteristic wavelengths. Because only electronic transitions are involved, the resulting emission lines are very narrow. The intensity...
244
Survival Tree
84
Survival trees are a non-parametric method used in survival analysis to model the relationship between a set of covariates and the time until an event of interest occurs, often referred to as the "time-to-event" or "survival time." This method is particularly useful when dealing with censored data, where the event has not occurred for some individuals by the end of the study period, or when the exact time of the event is unknown.
Building a Survival Tree
Constructing a...
Building a Survival Tree
Constructing a...
84


