Qing Liu

Rapid and accurate classification of surface defects in the strip steel production process contributes to reducing production costs in factories. In this article, a dual feature-integration network (DFINet) is constructed for faster and more pragmatic few-shot strip steel surface defect classification scenarios. The model utilizes a relatively simple convolutional neural network (CNN) as the backbone to improve real-time performance. Bidirectional long short-term memory networks are employed to extract cross-image features (CIFs) from the support set, which are, then, fused with the structural features extracted by the backbone to obtain aggregated features for updating deeper support set features. The positions of support samples are adjusted by randomly cropping them into several subimages and fusing the features of these subimages to further update the positions of support samples in the metric space. Leveraging the nonparametric structure and strong adaptability of Euclidean distance, the model employs it as a classifier to further enhance real-time and classification performance. Additionally, to simulate more pragmatic few-shot strip steel surface defect classification scenarios, a modification is made to the common N-way, K-shot training mode, by changing the number of samples in the support set from fixed-shot to Random-shot. Extensive experiments demonstrate that the proposed method can rapidly and accurately classify surface defects in strip steel, even in more challenging few-shot classification scenarios.

Strip steel plays a crucial role in modern industrial production, where enhancing the accuracy and real-time capabilities of surface defect classification is essential. However, acquiring and annotating defect samples for training deep learning models are challenging, further complicated by the presence of redundant information in these samples. These issues hinder the classification of strip steel surface defects. To address these challenges, this paper introduces a high real-time network, ODNet (Orthogonal Decomposition Network), designed for few-shot strip steel surface defect classification. ODNet utilizes ResNet as its backbone and incorporates orthogonal decomposition technology to reduce the feature redundancies. Furthermore, it integrates skip connection to preserve essential correlation information in the samples, preventing excessive elimination. The model optimizes the parameter efficiency by employing Euclidean distance as the classifier. The orthogonal decomposition not only helps reduce redundant image information but also ensures compatibility with the Euclidean distance requirement for orthogonal input. Extensive experiments conducted on the FSC-20 benchmark demonstrate that ODNet achieves superior real-time performance, accuracy, and generalization compared to alternative methods, effectively addressing the challenges of few-shot strip steel surface defect classification.

Circularly coded targets are widely used in 3D measurement, target tracking, augmented reality, and other fields as feature points to be measured. The traditional coded target recognition algorithm is easily affected by illumination changes and excessive shooting angles, and the recognition accuracy is significantly reduced. Therefore, a new coded target recognition algorithm is required to reduce the effects of illumination and angle on the recognition process. The influence of illumination on the recognition of coding targets was analyzed in depth, and the advantages and disadvantages of traditional algorithms are discussed. A new adaptive threshold image segmentation method was designed, which, in contrast to traditional algorithms, incorporates the feature information of coding targets in the determination of the image segmentation threshold. The experimental results show that this method significantly reduces the influence of illumination variations and cluttered backgrounds on image segmentation. Similarly, the influence of different angles on the recognition process of coding targets was studied. The coding target is decoded by radial sampling of the dense point network, which can effectively reduce the influence of angle on the recognition process and improve the recognition accuracy of coding targets and the robustness of the algorithm. In addition, further experiments verified that the proposed detection and recognition algorithm can better extract and identify with high positioning accuracy and decoding success rate. It can achieve accurate positioning even in complex environments and meet the needs of industrial measurements.

Aiming at the problem of constraints between reconstruction range and splicing accuracy when 3D reconstruction is carried out on large-scale equipment in aerospace and other fields, we investigate the combined 3D scanning measurement method based on the guidance of the all-space laser positioning system (ALPS) on the basis of analyzing the fusion of the ALPS and scanning target. The method establishes a global 3D scanning measurement model based on multiple measurement points by Creo and simulates and analyzes the data with MATLAB, then designs the structure of the circumferential combined photoelectric receiver by combining with the receiving characteristics of photodetectors for laser signals in the full-space field, and then completes the optimal layout design of the receiver based on the accessibility. Finally, a 30 m×30 m experimental platform for the fusion of ALPS and scanning target is constructed, and the measured standard bat and Zeiss standard ball are used as the measured objects to verify the feasibility of the research method. The simulation and experimental results show that the optimal number of receivers needed for the global measurement system is N=6, the measurement accuracy is better than 0.5 mm, and the repeatability measurement accuracy is up to 0.02 mm, which provides a basis for the 3D scanning of large-scale equipment.

The accurate large-scale positioning system is a novel full-space laser positioning system that has the advantages of automatic and multi-node parallel measurement and sub-millimeter positioning accuracy, and it has become an important component in the digital manufacturing of large equipment. Goniometric accuracy is particularly important for the evaluation of system function and accuracy improvement. An improved goniometric accuracy calibration method is proposed to assess the goniometric accuracy by using multiple measurement points to fit the mapping relationship between the goniometric measurement of the system and the goniometric measurement of the multi-tooth indexing table. The feasibility of the accuracy assessment method is verified by establishing a geometric model and performing modeling and simulation analysis. The method is verified using an all-space laser positioning system, and the dynamic performance of the base station is optimized to improve the overall goniometric accuracy. The maximum value of the goniometric error is 0.008″ and the goniometric accuracy is 0.026″ after fitting the curve by the mapping relation, demonstrating high stability, and the fluctuation of the goniometric error after optimization is improved from the original ±15″ to ±7.5″, verifying the effectiveness of the goniometric compensation method. This method solves the problem of coaxiality calibration in goniometric evaluation and provides the basis for the traceability of the goniometric accuracy.

Distributed laser measurement systems, widely used in high-end equipment such as airplanes, ships, and other manufacturing fields, face challenges in large spatial measurements due to laser plane obstructions and weak intersections. This paper introduces a novel omnidirectional sensor with enhanced adaptability to complex environments and improved measurement accuracy. Initially, an integrated omnidirectional measurement model is established, followed by the analysis of the optical path of the front-end detector, and the design of a signal-conditioning circuit for the photoelectric conversion of the front-end laser signal, Subsequently, a circuit testing platform is established to validate the detection functionality, and the corresponding results indicate that the symmetry of the output waveform is under 10 ns, the response time is under 100 ns, and the maximum detection distance is 22 m. Further, experimental results demonstrate the superiority of omnidirectional sensors over planar ones in complex environments, successfully receiving 360° laser signals. The positional accuracy of the common point to be measured on the top of the omnidirectional sensor is confirmed to exceed 0.05 mm, and the accuracy of the angle of attitude exceeds 0.04°. Using the laser tracker, the measurement accuracy of the system is verified to be better than 0.3 mm. When rotating in the horizontal and pitch directions, the measurement accuracy is better than 0.35 mm and 0.47 mm, respectively, fulfilling the sub-millimeter precision requirement and expanding the application scope of distributed laser measurement systems.

Aiming at the distributed measurement system multi-equipment coordinated measurement, the data measured by each transmitting station are unified under a common coordinate system, which has the problems of complicated operation and high time and cost. In this paper, a fast calibration method for distributed measurement systems based on standard devices is proposed. The method designs a new type of architectural calibration board with non-coplanar measurement nodes, calculates the coordinates of the measurement nodes using geometric relations, solves the initial values of the system parameters, and substitutes the initial values into the L-M least squares algorithm for iterative optimization so that it only needs to carry out one scanning measurement for each reference position in the calibration process to complete the calibration. We designed an experiment in which a distributed measurement system consisting of four transmitting stations measured three standard rods (1650.073, 1156.704, and 875.950 mm) sequentially after calibration in a measuring space of 5 × 5 × 3 m3. The experimental results show that under the premise of ensuring that the measurement accuracy is better than 0.20 mm, the calibration of the parameters of the distributed measurement system can be completed within 20 min.