激光与光电子学进展, 2017, 54 (8): 081005, 网络出版: 2017-08-02
基于多特征融合的玉米前期图像的旱情识别
Drought Identification Based on Multi-Features Fusion for Early Maize Images
摘要
为实现对玉米植株旱情的分析, 针对目前农业干旱指标涉及领域较为广泛、获取困难的研究现状, 提出了一种基于多特征融合的玉米前期图像旱情识别方法。以正常和特旱两种情况的玉米植株图像为样本, 采用经典K-means算法对玉米植株图像提取感兴趣区域;进而提取分割后的玉米植株图像, 包括颜色、奇异值分解(SVD)、纹理等共计20维特征;采用遗传算法对20维特征选择有效特征子集;最后针对有效特征子集建立了基于最小二乘支持向量机的判别模型, 获取了玉米植株图像的旱情信息。将单个特征(颜色、SVD、纹理)直接融合之后的特征以及利用主成分分析法的特征选择作为对比实验, 平均识别正确率分别为 0.9503、0.9627、0.9771、0.9460、0.9745, 而采用遗传算法进行特征选择后, 最终寻到最优解为9维特征, 平均识别正确率为0.9903。结果表明,运用图像处理技术可以对旱情进行识别, 取得了较好的效果, 为农业旱情的识别提供了新思路。
Abstract
In order to analyze the drought of maize plants and aiming at the difficulty and broad in recognizing agricultural drought, we propose a method to identify the drought of maize plants based on the multi-features fusion. The images of normal and seriously drought plants are taken as samples. The K-means algorithm is used to extract the interesting areas of maize plant images. And, the features of the pictures are extracted after image segmentation, including colors, singular value decomposition (SVD) and textures, a total of 20 dimensional features. The genetic algorithm is used to select a effective features subset of 20 dimensional features. Finally, the discrimination model based on least squares support vector machine is established for the effective features subset and images of maize plant drought are obtained. The single feature (color、SVD、texture) after directly fusion and using principal component analysis for feature selection are performed as comparative experiments, the average recognition accuracies are 0.9503, 0.9627, 0.9771, 0.9460, 0.9745, respectively. The genetic algorithm is used to select the features, and finally finds 9 dimensional features as the optimal solution. The average recognition accuracy is 0.9903. The result shows that this image processing technology can identify the drought situation of the maize plants effectively and efficiently. And it also provides a new idea for the drought identification of maize plants.
路志英, 刘书辰, 宫志宏. 基于多特征融合的玉米前期图像的旱情识别[J]. 激光与光电子学进展, 2017, 54(8): 081005. Lu Zhiying, Liu Shuchen, Gong Zhihong. Drought Identification Based on Multi-Features Fusion for Early Maize Images[J]. Laser & Optoelectronics Progress, 2017, 54(8): 081005.