摘要
由于现有的用于农业作物生长监测数据的特征优化方法—局部保持投影(locality preserving projection,LPP)只保留局部信息,同时存在未考虑样本类别信息导致特征提取时误分类,准确率与数据优化效率并不理想。针对上述问题,提出了改进型LPP方法,并将其用于作物生长特征的优化。首先将样本利用二维主成分分析(two-dimensional principal component analysis,2DPCA)进行初步降维,保留原样本数据中的整体空间信息;然后提出优化的2类子图—聚集子图和分离子图,用来描述不同类别数据之间的关联信息;然后提出优化的2类子图对不同类别数据间的远近关系进行描述;最后采用改进型LPP算法,将数据进一步投影到低维空间,提取样本的局部信息,完成样本特征优化。试验分析表明,改进型LPP具有很好的适应性,最高支持向量机(support vector machine,SVM)分类准确率能够达到96%以上,使精度达到最高的最优维数比主成分分析(principal component analysis,PCA)和二维主成分分析2种算法降低了25%以上,同时算法运行效率比PCA与2DPCA算法提升32.4%与8.3%,整体性能比基本LPP算法更为优越,能够适应农作物多维数据的优化处理。研究结果为现代精准农业信息监测过程中的数据处理与分析提供了参考。
Nowadays, the evaluation for crop growth is based on various growth characteristics, which often brings a huge amount of information processing. Furthermore, the complex information can not directly reflect some key features of crops. Thus, the feature extraction and optimization plays an important role in the process. In this paper, the locality preserving projection (LPP) is used to achieve the dimensionality reduction of high dimensional data while keeping the invariance of its internal local structure. After being projected via the algorithm, the adjacent sample is able to maintain the original neighboring state while the original distant samples don’t keep the old state. Obviously, this result is not satisfactory for data optimization. In order to strengthen the effect of category separation, firstly, the dimension of sample data is preliminary reduced by using two-dimensional principal component analysis (2DPCA) to retain the spatial information. Secondly, two sub-graphs (gathered sub-graph and separated sub-graph) instead of original nearest neighbor graph are used to describe the relationship between homogeneous and heterogeneous data. The gathered sub-graph is improved on the basis of K-nearest neighbor graph. The addition of category information makes the homogeneous non-adjacent samples stay closer to each other after projection. The separated sub-graph is constructed to solve the problem that the application of K-nearest neighbor graph may reduce the accuracy of classification when the data are projected into low-dimensional space. Then the optimized global matrix and the improved objective function are provided to design the complete optimization method for feature extraction. Through the above steps, the category information of sample data is added for LPP algorithm. Finally, the feature parameters set are obtained by improved LPP algorithm to extract local information of samples. The data of crop growth features are further projected to low-dimensional space. The final extracted information is able to replace the original sample data without losing the data which can reflect the key information of sample set. In order to evaluate the performance of improved LPP algorithm to achieve dimensionality reduction and optimizing for crop growth characteristics, a set of data from cabbage was chosen as test sample. In the process of dimensionality reduction from 30 to 10 using different algorithms (PCA, 2DPCA, LPP and improved LPP), the improved LPP has higher overall performance with less running time, which is only longer than Basic LPP algorithm. By analyzing the performance of improved LPP algorithm for dimensionality reduction, the data of some cabbage and lettuce were chosen as test data. The contrast experiments using different algorithms (PCA, 2DPCA, LPP and improved LPP) for dimensionality reduction were carried out, and all the test data in the database achieved dimensionality reduction via the above-mentioned algorithms. Meanwhile, it accomplished data classification by SVM after accomplishing dimensionality reduction. The experiments show that the improved LPP algorithm has better adaptability, and the highest SVM classification accuracy rate of this method can reach up to 96%. Compared with other methods, the improved LPP has superior performances in terms of multidimensional data analysis and optimization. The method has good prospects, and is able to meet the demands for the information perception of new agriculture as well as the optimization of crop growth characteristic parameters.
出处
《农业工程学报》
EI
CAS
CSCD
北大核心
2014年第15期206-213,共8页
Transactions of the Chinese Society of Agricultural Engineering
基金
国家发改委资助智慧农业示范项目(W13DR00020)
关键词
优化
数据处理
监测
生长特征
特征优化
局部保持投影
二维主成分分析
optimization
data processing
monitoring
growth features
feature optimization
locality preserving projection
two-dimensional principal component analysis