| 摘要: |
| 【目的】针对寒地大豆发生缺素症状时冠层颜色变化复杂性，建立基于冠层图像颜色特征的大豆缺素症状识别新方法。【方法】采用无土盆栽试验，以垦农18为供试大豆品种，设计缺氮、缺磷、缺钾3种营养状况，采集大豆缺素症状的冠层图像样本，利用图像灰度直方图结合主成分分析方法,提取大豆冠层图像的红光值R、绿光值G、蓝光值B，计算最佳颜色特征蓝光标准化值B/(R+G+B)和绿光标准化值G/(R+G+B)，将其作为正则化模糊神经网络输入向量，并利用实数编码的遗传算法改进传统梯度下降学习算法，将其作为模糊神经网络的学习方法，同时应用传统梯度下降算法和改进梯度下降算法训练神经网络参数并比较。【结果】应用遗传计算改进的梯度下降学习算法计算时，迭代次数为277次，其各项计算指标均明显优于传统梯度下降算法，大豆缺素症状识别准确率达100%；而采用传统的多元线性回归方程和BP神经网络算法计算时，识别准确率分别为52.50%，68.33%。【结论】以大豆冠层图像颜色特征为基础，利用改进学习算法的神经网络模型，能够快速有效地挖掘出大豆缺素症状与颜色特征向量之间的模糊逻辑映射关系，为大豆缺素症状识别提供了一种快速且准确的方法。 |
| 关键词: 冠层图像 颜色特征 大豆缺素症状 识别方法 |
| DOI： |
| 分类号: |
| 基金项目:黑龙江省自然科学基金项目（QC2016031）；黑龙江省大学生创新创业训练计划项目(1022320169433)；黑龙江省教育厅科学技术研究项目（12521375） |
|
| Recognition of soybean nutrient deficiency based on color characteristics of canopy |
|
GUAN Haiou,LI Jiapeng,MA Xiaodan,et al
|
| Abstract: |
| 【Objective】Aiming at the complexity of color spatial characteristics of soybean canopy with nutrient deficiency in cold area,a new method based on canopy color characteristics was established to recognize the soybean nutrient deficiency.【Method】Based on soilless potting,three nutrient conditions of nitrogen deficiency,phosphorus deficiency,and potassium deficiency were applied to soybean variety Kennong 18.Image histogram combined with principal component analysis method was used to extract standardization values of blue B/(R+G+B) and green G/(R+G+B),which were used as vector bases of automatic reasoning for regularized fuzzy neural network.Then,gradient descent algorithm improved by genetic algorithm was adopted as neural network learning method,which was compared to traditional gradient descent algorithm in parameters of neural network.【Result】The learning number of the established method was 277,other indexes are all better than that of tradition gradient descent algorithm the recognition accuracy rate was close to 100%,while the accuracy rate of traditional multiple linear regression equation and BP neural network are 52.50% and 68.33% respectirely.【Conclusion】The relationship between nutrient deficiency of soybean and corresponding color characteristic vector can be extracted by the established method in this paper.It provides a new calculation method to recognize soybean nutrient deficiency fast and accurately. |
| Key words: canopy image color characteristic soybean nutrient deficiency recognition method |
