基于动态高分辨图像的粮仓玉米温度变化监测方法  

Dynamic high-resolution images-based monitoring of corn temperature in granary

作　　者：张文静[1] ZHANG Wen-jing(School of Intelligent Agricultural Engineering,Beijing Vocational College of Agriculture,Beijing 102442,China)

机构地区：[1]北京农业职业学院智慧农业工程学院,北京102442

出　　处：《粮食与饲料工业》2024年第6期15-18,23,共5页Cereal & Feed Industry

摘　　要：为了监测粮仓内玉米的温度变化情况,为后续的储藏条件提供理论参考,提出了一种基于动态高分辨率图像的玉米粮仓温度变化监测方法。采用热成像传感器进行玉米粮仓基础数据采集,数据由H3C R4900 G3服务器处理和存储。运用绝对定标方法计算干扰率以消除传感器干扰,生成高分辨率图像,并进行几何校正。然后通过影像分割提取差异性特征,影像分割将图像划分为不同区域后提取纹理特征确定监测范围,经多尺度分割和区域整合得到影像像斑,依据像斑差异性参数求解公式确定像斑整合条件。最后在多维特征下进行动态高分辨率监测玉米粮仓温度变化,通过设定卷积滤波中心、卷积层特征表达式等一系列计算得到影像像斑特征,从而实现温度变化监测。实验结果表明:该方法能够有效地监测粮仓玉米温度变化,为后续的粮仓存储提供了有价值的参考。In order to keep track of temperature of corn stored in granary and provide theoretical reference for subsequent storage conditions,this paper proposed a temperature monitoring method for stored corn with dynamic high-resolution images as the basis.Fundamental data were first collected from corn granary using thermal imaging sensors and then processed and stored in H3C R4900 G3 server.Absolute calibration method was employed to determine interference rate and eliminate interference from sensors before generation of high-resolution images.The images received further geometric correction.To extract differential characteristics,image segmentation was applied to divide images into several regions for extracting textural features and determining monitoring scope.After multi-scale segmentation and regional integration,image spots were obtained.Image spot integration conditions were figured out by solving formulas with image spot differential parameter.Finally,dynamic high-resolution monitoring was achieved on temperature change within corn granary under multidimensional features.By computing convolution filtering center and convolution layer characteristics,image spot features were summarized for successfully monitoring temperature change.The experimental results indicated the proposed method could effectively keep track of the temperature within corn granary and provide valuable reference for future warehousing.

关 键 词：玉米粮仓 热成像传感器 高分辨率影像 多维特征 卷积神经网络 

分 类 号：S379[农业科学—农产品加工] TP183[农业科学—农艺学]