基于Gabor变换和极限学习机的贝类图像种类识别

Shellfish recognition based on Gabor transformation and extreme learning machine

研究了基于Gabor变换和二维图像主成分分析(2DPCA)相结合的贝类图像识别方法。对贝类图像进行Gabor变换,提取其图像特征,确定了图像特征维数;采用2DPCA方法,对变换后的特征进行降维,并利用极限学习机(ELM)进行贝类图像的分类识别。与BP神经网络和支持向量机(SVM)实验对比发现,极限学习机分类器用于贝类识别不仅速度极快而且泛化性良好,算法具有较高的精度。

A method identifying shellfish image is proposed based on Gabor transformation and 2-dimensional principal component analysis （2DPCA）. The characteristic was firstly extracted from shellfish image and the dimension magnitude of image is confirmed using Gabor transform. 2DPCA method was then employed to reduce the dimension of transformed characteristic. And the extreme learning machine （ELM） is employed to conduct the classification and recognition of shellfish image. Compared of back propagation neural network with support vector machine （SVM） experimental technique showed that ELM classifier has good speed and applicability for shellfish recognition, and the method was proved of higher accuracy and efficiency.