面向自动拆解的废旧电器整机智能识别方法

针对拆解废旧电器整机识别的传统方法效率低下的现象,提出一种自定义特征的废旧电器整机识别的方法;首先对废旧电器图像采用目标分割算法把废旧电器与背景进行分割,然后提取废旧电器整机的形状特征和卷积神经网络提取的深层特征,采用PCA算法对提取到的形状特征进行优化,将优化后的形状特征与深层特征进行特征拼接,最后将拼接后的特征向量对搭建好的3个SVM二分类器进行训练,得到废旧电器的分类模型;结果表明,拼接后的特征向量对废旧电器识别的准确率较高,高达91.21%,能够有效地实现废旧电器的智能识别。

Intelligent identification method of waste electrical equipment for automatic disassembly

In view of the inefficiency of the traditional identification method of disassembled waste electrical equipment, a method of identifying waste electrical equipment with custom features was proposed. Firstly, the image of waste electrical appliances was segmtioned from the background by object segmentation algorithm. Then, the shape features of the waste electrical appliances and the deep features extracted by convolutional neural network were extracted. PCA algorithm was used to optimize the extracted shape features, and the optimized shape features were splice with the deep features. Finally, the spliced feature vectors are trained on the three SVM binary classifiers, and the classification model of waste electrical appliances is obtained. The results show that the recognition accuracy of the splice feature vector is high, up to 91.21%, which can effectively realize the intelligent identification of waste electrical appliances.