【作者】 关海鸥；

【导师】 许少华；

【作者基本信息】 大庆石油学院 ， 计算机软件与理论， 2010， 硕士

【摘要】 本文对植物叶片非线性形变的病害图像的病斑识别与特征计算技术进行研究,利用图像处理、神经网络及植物病理学知识相融合,实现了对非线性失真的植物叶片的数字图像进行校正,病害区域识别,病斑区域特征的提取和计算。针对植物叶片图像的非线性失真形式,提出双线性映射方程来恢复失真图像原来的空间关系,利用最近邻插值法进行图像的灰度插值,完成对失真图像的校正。利用植物病斑区域识别分割的技术,提出了在输出层采用线性清晰化方式的五层的模糊神经网络模糊推理系统。采用梯度下降的误差反向传播法进行参数辨识,针对BP算法学习收敛速度慢且易陷入局部极小的不足,提出一种遗传算法和梯度下降的误差反向传播法相融合技术的学习算法,充分利用了遗传算法的全局搜索最优性和梯度下降的误差反向传播局部搜索最优性有机结合一种较优秀的网络学习算法。依据植物病害智能诊断的条件因素,利用图像平滑技术的滤波方法对病斑图像完成预处理后,能够实现对几何、颜色、纹理诊断依据的特征提取和计算。针对其特征矢量处于特征空间,提出隐含层的量子神经元为反正切函数的三层量子神经网络的植物病害诊断模型,样本数据中的不确定性在给定梯度下降学习算法的训练过程中被自适应的确定。本研究的植物病害无损智能诊断关键技术在马铃薯早疫病的诊断中取得了显著效果。更多还原

【Abstract】 This paper studies the crucial technologies about intellectual diagnosis of diseases of plants and realizes correction of digital images, recognition of diseased area, the extraction and calculation of diseased area’s features,according to image processing, neural net and plant pathology.Considering the nonlinear distortion of images of leaves, bilinear map equation is proposed to restore distorted images by utilizing interpolation, so that the correction of distorted images is obtained. Meanwhile, the five-layer fuzzy inference system is given in output layer by the partition of diseased area. Because BP algorithm has a slow convergence and usually is trapped into the local minimum, a new algorithm is proposed, which is a more excellent one that is combining two algorithms: genetic algorithm and BP. The extraction and calculation of features in geometry, color and texture are achieved after the pre-processing of the image with scabs by utilizing smoothing technique according to the factors of intellectual diagnosis for diseased plants. Besides these above, a model is designed that is a three-layer quantum neural network model, whose hidden layer is arctangent function. By the improved model, uncertainty of sample data can be determined during the training process in the given gradient descent algorithm. The results of stimulation indicate that the crucial technology proposed in the paper can achieve remarkable recognition and judgment in diagnosis of potatos’early blight.更多还原