Research on Multi-side Joint Industrial Parts Inspection Based on Model Fusion. Issue 1 (31st December 2022)

• Record Type: Journal Article
• Authors: Jiao, Lei; Lin, Xuesong; An, Kang; Song, Yaqing; Liu, Yingyuan; Liu, Hui; Nan, Hai
• Is Part Of: Applied artificial intelligence. Volume 36:Issue 1(2022)
• Publication Date: 2022-12-31
• Subjects: Artificial intelligence -- Periodicals; 006.3
• DOI: 10.1080/08839514.2022.2055396
• Languages: English
• ISSNs: 0883-9514

Abstract

ABSTRACT: Rapid identification of many industrial material parts is the key to effectively improving the efficiency of the industrial production process and intelligent warehousing. Accurate identification of both similar and special parts is an important problem. In this paper, a multi-side joint industrial part recognition method based on model fusion learning is proposed. A multi-channel visual acquisition system is designed to construct a fast industrial part data set with time and space consistency. Two joint identification methods for multi-side acquisition are proposed, and the classification results are fused to improve the model's accuracy and solve the classification problem for similar parts. The experimental results show that compared with the traditional model, the prediction accuracy of the multi-channel multi-input model proposed in this paper is improved by approximately 6%, and the accuracy of the single-channel multi-input model is improved by approximately 10%. The accuracy of part recognition is far better than that of the traditional model, and it therefore provides a new strategy for the rapid training and recognition of industri-al parts in intelligent storage.