CAJ | 학술논문

为了对车载玉米叶绿素含量快速预测系统偏差进行分析,优化车载系统测量结果,该文提出一种空间插值和多维矩阵分析相结合的思想,阐述了基于矩阵变换和GIS空间分析手段的预测偏差分析策略,提高了车载系统快速预测空间分布的预测效果.研究结果表明:使用BP神经网络模型和RBF神经网络模型对车载系统动态预测单点位置叶绿素含量具有一定预测效果,平均决定系数R2约为0.8,2类模型的预测效果相近.RBF神经网络预测结果矩阵经反距离加权法插值后,其空间分布预测偏差度小于10%的数据量占总数据量的85%.表明该方法具有较好的空间预测效果,可以为车载系统动态测量平台预测玉米冠层叶片叶绿素含量的提供决策支持.
위료대차재옥미협록소함량쾌속예측계통편차진행분석,우화차재계통측량결과,해문제출일충공간삽치화다유구진분석상결합적사상,천술료기우구진변환화GIS공간분석수단적예측편차분석책략,제고료차재계통쾌속예측공간분포적예측효과.연구결과표명:사용BP신경망락모형화RBF신경망락모형대차재계통동태예측단점위치협록소함량구유일정예측효과,평균결정계수R2약위0.8,2류모형적예측효과상근.RBF신경망락예측결과구진경반거리가권법삽치후,기공간분포예측편차도소우10%적수거량점총수거량적85%.표명해방법구유교호적공간예측효과,가이위차재계통동태측량평태예측옥미관층협편협록소함량적제공결책지지.
@@@@Using near-ground remote sensing is significant to understanding the growth of crops and providing accurate and scientific data for precision agriculture. The vehicle-borne system is one of the most important tools for growth monitoring and management. It is an efficient, flexible and economical operation for the small region. However, the vehicle-borne growth monitoring system cannot maintain steady operations due to the row spacing of corn. The background interference on the reflectance will not be suppressed effectively, which will result in a deviation in the growth monitoring. In order to overcome this problem, a new method was developed in this paper, which contains matrix transformation and GIS analysis. @@@@In order to obtain the experimental data, the tests were carried out by the vehicle-borne system on the cornfield. The vehicle-borne system collected the reflectance data of the corn canopy with the sensors at a sampling rate of 1 point per second. The GPS receiver obtained the location information as the same rate. All information was formulated in a matrix at each experiment. Then, each data set of the matrix was combined by the information of GPS and canopy reflectance. The spatial interpolation methods of Inverse Distance Weighted (IDW) and Kriging were utilized for comparison study on the matrix. It overcomes the shortcomings of the large deviation resulting for the background interference. By dealing with neural network analysis between reflectance and chlorophyll content, the results analysis from the matrix can show the corn growth in some specified region. @@@@The results indicated that:It has satisfactory forecasting accuracy on the chlorophyll content by using the BP neural network model and RBF neural network model, with average R2 of 0.8. By focusing on the optimization of the spatial data distribution obtained by vehicle, it was proposed that the matrix of results, which was predicted by RBF neural network, was transformed with inverse distance weighted (IDW). It was under 10% that the deviation rate between the predication and real value was the majority, which was about 85%of the entire data. @@@@The theoretical analysis and test results prove that the method of combining spatial analysis, neural network and matrix transformation has the characteristics of estimating the corn growth by the traverse measurement system. It also showed the good effect on solving the dynamic crop growth predication with severe background interference.