Key points are not available for this paper at this time.
Industry 4.0 aims to address the issues of low accuracy and efficiency in the identification and positioning of components during machine tool processing and assembly. To this end, a novel component identification and positioning algorithm, LAI YOLOv5, has been proposed. This algorithm integrates lightweight networks, attention mechanisms, and information fusion techniques. Initially, the convolution layers in the YOLOv5 network structure are optimized for lightweight processing, effectively reducing the number of neural network parameters and floating-point operations, thereby decreasing memory usage and enhancing real-time detection speed. Subsequently, an attention mechanism is introduced into the backbone network to improve the specificity of feature extraction and enhance the salience of detected objects. Finally, a cross-channel information fusion mechanism is incorporated into the feature fusion network to boost feature detection capabilities. Experimental results indicate that compared to the original algorithm, the improved LAI YOLOv5 algorithm reduces the number of parameters and network layers by approximately 45.98% and 28.46%, respectively, decreases the floating-point operations by about 55.82%, reduces memory usage by 15.51%, and shortens training time by around 32.27%. Additionally, the training accuracy reaches 96.80%, training coverage reaches 95.01%, real-time detection efficiency improves to 100.739 FPS, and detection accuracy achieves 98.62%.
Hong et al. (Sat,) studied this question.