Abstract:  In order to further improve the performance of detection and recognition algorithm for navel orange disease, a kind of fuzzy detection method based on the watershed segmented edge self-organizing neural network(FSONN) is proposed. Firstly, the watershed segmentation algorithm was used to realize the effective extraction of the boundary of navel orange, and used the expression of red, green components in the disease region to characterize the disease of navel orange, used the fractal dimension of the navel orange disease to be the shape expression, and used the above features as the input of neural network algorithm to construct an automatic detection model of navel orange disease. Secondly, the fuzzy self-organizing neural network algorithm was proposed to realize the automatic model parameters and structure identification based on the self-organizing clustering method; Finally, through the experiment on the MackeyGLass approximation of nonlinear sequences in the proposed algorithm, the performance advantages can be verified, and applies it in navel orange disease detection, the experimental results show that the correct rate of 4 kinds of navel orange disease detection by proposed algorithm can reach more than 90%, which can meet the needs of practical applications.

Key words: Watershed segmentation algorithm, edge detection, navel orange disease, fuzzy self-organizing neural network

摘要： 为进一步提高脐橙病害检测识别算法性能，提出一种基于分水岭分割边缘自组织神经网络（Fuzzy self organizing neural network, FSONN）的脐橙病害模糊检测方法。首先，基于分水岭分割算法实现脐橙病害边界的有效提取，以病害区的红、绿颜色分量对脐橙病害进行特征表达，并以分形维数对脐橙病害形状进行特征表达，然后将上述特征作为神经网络算法的输入，构建脐橙病害自动检测模型；其次，引入模糊自组织神经网络算法，实现基于自组织聚类方式的自动模型参数和结构辨识；最后，通过在MackeyGLass非线性序列中的逼近实验验证了所提算法性能优势，并将其应用于脐橙病害检测，实验结果表明，所提算法的4种脐橙病害检测正确率可达到90 %以上，满足实际应用需要。

关键词: 分水岭分割算法, 边缘检测, 脐橙病害, 模糊自组织神经网络