基于改进极限学习机的透明塑料裂痕检测

Transparent Plastic Crack Detection Based on Improved Extreme Learning Machine

针对透明塑料中微小裂痕难以检测的问题,提出了基于改进的极限学习机算法的检测方法,采用卷积神经网络以组建特征提取器;同时,采用基于狮群算法优化的改进极限学习机算法以构建分类器。在改进极限学习机算法中,狮群算法被用于优化隐含层神经元和输入层神经元之间的权重矩阵,提高了透明塑料微小裂痕检测实验中的识别率。

Aiming the micro cracks in transparent plastic, which are difficult to detect, this paper proposes a detection method based on improved extreme learning machine algorithm. The convolutional neural network is used to construct the feature extractor and the improved extreme learning machine algorithm based on the lion group algorithm is used to construct the classifier. In the improved extreme learning machine algorithm, the lion group algorithm is used to optimize the weight matrix between the hidden layer neurons and the input layer neurons. The detection method proposed in this paper achieves higher recognition rate than the traditional method in the transparent plastic micro-crack detection experiment.