Greengage grading using stochastic configuration networks and a semi-supervised feedback mechanism.
Information Sciences(2019)
摘要
In order to gain a competitive advantage in the international market, the automatic and accurate grading of fruit and vegetables, which is grouping fruit by either size, color or other biological characteristics is crucial to ensure quality. To minimize the subjectivity of recommendations made in relation to grading fruit when this is done by many individuals, a greengage grading model with stochastic configuration networks (SCNs) and a semi-supervised feedback mechanism is proposed in this paper. In this model, the semi-supervised mechanism is firstly employed to expand the training dataset by labelling the untagged images to improve the performance of existing supervisory models. Then, to facilitate the building of the learner model, a compact set of features that possesses a sufficient amount of information and the discriminative power can be extracted from the training dataset using adaptive convolutional neural networks (ACNNs) is sent to a SCN learner with a universal approximation ability, and an alternating optimization technique is applied to update the ACNN model and generate a new SCN model. Finally, according to the constraints of the semantic error entropy measure, uncertain grade outputs of the testing greengage images are evaluated in real time to update the multilevel knowledge space with feedback and a self-optimization mechanism. Comprehensive simulation results indicate the merits of our proposed model in terms of accuracy and robustness compared to the other open-loop and closed-loop supervisory methods.
更多查看译文
关键词
Greengage grade,Stochastic configuration networks,Semi-supervised,Convolutional neural network,Semantic error entropy
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络