作　　者： ; ; ;

机构地区： 仲恺农业工程学院

出　　处： 《农机化研究》 2015年第2期171-175,共5页

摘　　要： 为探索更实用的叶面积地面测量方法,以TSL2561为核心研制了一款数字式光强传感器阵列,制作了升降式平台进行试验。试验系统硬件采用"PC机-单片机"分布式结构,运用MatLab语言编写上位机数据采集和叶面积计算软件。在试验平台上3个不同的置物高度进行了100个细叶榕叶片样本的面积测定,分别运用多元线性回归、BP神经网络和支持向量机3种算法建立光强数据与叶面积的关系模型。结果表明:扫描图像的分辨率、目标形状对叶面积图像测算法得出的结果影响不显著;最佳置物高度为30cm;支持向量机算法效果最好,决定系数R2为0.973 3,平均相对误差为3.70%,均方根误差为0.587 2。此模型在系统中的运用可直接、快捷地测算出叶面积数据,为利用多波段透射光技术测量单叶片的叶面积提供了参考。 A digital light sensor array and a 100 cm high elevator-platform were developed for the purpose of seeking for a more practical method to measure leaf area. 9 chips of TSL2561 were used to constitute the sensor array developed in flat surface structure on a piece of printed circuit board. A testing system based on distributed structure was established. Hardware of the system consisted of data collection terminal based on AT Mega128L, light sensor array and personal com- puter. Software of the system, which was complied by Matlab, had the abilities to collect data and calculate leaf area. 100 leaf samples of ficus microcarpa were plucked randomly and the area of each leaf sample was measured at three dif- ferent heights using the testing system. A reference system based on existing image scanning method was set up to meas- ure and calculate the leaf area of each sample. Regression methods were used to establish models for calculating leaf area based on data collected from the sensor array and image scanning system. Relational models were setup and reached their best effect when the placing table was at the height of 30 era. Multiple relational models were established using different methods such as liner, BP neural network and support vector machine（SVM）. According to the result of the test, res- olution ratio and target shape were not significant to the area calculation when image scanning method was used, SVM method was the best of three and the experiment on the test set reached a square correlation coefficient （R2） of 0. 9733, a mean relative error of 3.7% and a root mean square error （RMSE） of 0. 587 2. The result of the measure- ment for single leaf area can be directly and quickly obtained by applying this model.

关 键 词： 叶面积 传感器阵列 图像处理 测量平台

领　　域： [农业科学] [农业科学] [自动化与计算机技术] [自动化与计算机技术]