基于图像检测技术的室内人员服装热阻测量研究

Research on Measurement of Thermal Resistance of Indoor Personnel Clothing Based on Image Detection Technology

摘要:

目前服装热阻主要采用人工问卷调查的方式测量，需要受试者多次填写问卷，估计过程复杂且不易实时测量，而且传统估计方法只对静态热阻做预测，没有考虑人员运动状态、室内风速的影响。图像检测方面，基于Mask RCNN网络的服装检测方法存在多尺度特征信息丢失、融合不佳等问题。针对这些情况，提出一种改进的Mask RCNN服装检测网络方法，应用并实现室内人员动态服装热阻的系统设计。首先，通过CCD相机进行图像采集，经过改进的Mask RCNN网络检测着衣量。然后，查表映射法对室内服装热阻进行初步估计。最后，利用测量仪器测得风速、行走速度对服装热阻修正，得到动态服装热阻估计结果。实验结果表明，改进的Mask RCNN网络平均识别精度比原方法提高了1.1%，在动态服装热阻估计方面，与传统方法相比，能修正0.13的平均偏差。

Abstract:

At present, clothing thermal resistance is mainly measured by manual questionnaire, which requires the subjects to fill in the questionnaire many times. The estimation process is complex and not easy to measure in real time. Moreover, the traditional estimation method only predicts the static thermal resistance, and does not consider the influence of personnel movement state and indoor wind speed. In the aspect of image detection, the clothing detection method based on mask RCNN network has the problems of multi-scale feature information loss and poor fusion. In view of these situations, an improved mask RCNN clothing detection network method is proposed to apply and realize the system design of indoor personnel dynamic clothing thermal resistance. Firstly, the CCD camera is used for image acquisition, and the improved mask RCNN network is used to detect the amount of clothing. Then, look-up table mapping method is used to estimate the thermal resistance of indoor clothing. Finally, the wind speed and walking speed measured by the measuring instrument are used to correct the thermal resistance of clothing, and the estimation result of dynamic thermal resistance of clothing is obtained. The experimental results show that the average recognition accuracy of the improved mask RCNN network is 1.1% higher than that of the original method. Compared with the traditional method, the improved mask RCNN network can correct the average deviation of 0.13.