Wedge angle and orientation recognition of multi-opening objects using an attention-based CNN model

• DOI: 10.1364/oe.529655
• Publication Date: 2024-07-17T08:00:11Z
• Authors: Yiwen Zhang; Si-Ao Li; Xiaoyan Wang; Yongxiong Ren; Zihan Geng; Fei Yang; Zhongqi Pan; Yang Yue
• Keywords: Wedge (geometry); Feature (linguistics); Identification

In industries such as manufacturing and safety monitoring, accurately identifying the shape characteristics of multi-opening objects is essential for assembly, maintenance, fault diagnosis machinery components. Compared to traditional contact sensing methods, image-based feature recognition technology offers non-destructive assessment greater efficiency, holding significant practical value in these fields. Although convolutional neural networks (CNNs) have achieved remarkable success image classification tasks, they still face challenges dealing with subtle features complex backgrounds, especially similar openings, where minute angle differences are critical. To improve identification accuracy speed, this study introduces an efficient CNN model, ADSA-Net, which utilizes additive self-attention mechanism. When coupled active light source system, ADSA-Net enables non-contact, high-precision 14 classes rotationally symmetric multiple openings. Experimental results demonstrate that achieves accuracies 100%, ≥98.04%, ≥98.98% number wedge angles, opening orientations all objects, respectively a resolution 1°. By adopting linear layers replace quadratic matrix multiplication operations key-value interactions, significantly enhances computational efficiency accuracy.