The task of steel surface defect recognition is an industrial problem with great industry values. The data insufficiency is the major challenge in training a robust defect recognition network. Existing methods have investigated to enlarge the dataset by generating samples with generative models. However, their generation quality is still limited by the insufficiency of defect image samples. To this end, we propose Stable Surface Defect Generation (StableSDG), which transfers the vast generation distribution embedded in Stable Diffusion model for steel surface defect image generation. To tackle with the distinctive distribution gap between steel surface images and generated images of the diffusion model, we propose two processes. First, we align the distribution by adapting parameters of the diffusion model, adopted both in the token embedding space and network parameter space. Besides, in the generation process, we propose image-oriented generation rather than from pure Gaussian noises. We conduct extensive experiments on steel surface defect dataset, demonstrating state-of-the-art performance on generating high-quality samples and training recognition models, and both designed processes are significant for the performance. Note to Practitioners —This article introduces StableSDG, a method that generates realistic defect images even with limited data. It overcomes the shortcomings of current deep learning approaches that need large datasets to train from scratch. Our solution is to adapt a text-to-image diffusion model for defect generation. The proposed strategy involves two processes: training to adapt token embeddings and model parameters, and generation from partially perturbed defect images. The results show enhanced generation quality and improved accuracy for recognition models trained on the expanded dataset. StableSDG can be practically applied to efficiently enlarge a defect dataset, even when starting with a small amount of data.

Feature representation is the key to the hyperspectral images (HSI) inpainting. Existing works mainly focus on using spectral and temporal auxiliary images to inpainting the corrupted region, which were proved to be low robustness for all bands missing and high requirements for image acquisition. In this work, we propose an end-to-end inpainting framework for hyperspectral images based on convolutional neural networks, which does not require auxiliary images and makes full use of both spectral characteristics and spatial information. For spectral characteristics, a channel attention mechanism is proposed to reduce the redundancy of hyperspectral channels and model the correlation between channels. For spatial information, a local discriminative network is able to cope with the structural continuity of the corrupted regions, and a gradient consistency loss function is proposed to maintain the texture consistency of HSIs. Experimental results in the AVIRIS Indians Pines public dataset and FCH datasets show that our proposed method can provide competitive results compared with state-of-the-art methods.

High-precision and high-density three-dimensional point cloud models usually contain redundant data, which implies extra time and hardware costs in the subsequent data processing stage. To analyze and extract data more effectively, the point cloud must be simplified before data processing. Given that point cloud simplification must be sensitive to features to ensure that more valid information can be saved, in this paper, a new simplification algorithm for scattered point clouds with feature preservation, which can reduce the amount of data while retaining the features of data, is proposed. First, the Delaunay neighborhood of the point cloud is constructed, and then the edge points of the point cloud are extracted by the edge distribution characteristics of the point cloud. Second, the moving least-square method is used to obtain the normal vector of the point cloud and the valley ridge points of the model. Then, potential feature points are identified further and retained on the basis of the discrete gradient idea. Finally, non-feature points are extracted. Experimental results show that our method can be applied to models with different curvatures and effectively avoid the hole phenomenon in the simplification process. To further improve the robustness and anti-noise ability of the method, the neighborhood of the point cloud can be extended to multiple levels, and a balance between simplification speed and accuracy needs to be found.

Portable box volume measurement has always been a popular issue in the intelligent logistic industry. This work presents a portable system for box volume measurement that is based on line-structured light vision and deep learning. This system consists of a novel 2 × 2 laser line grid projector, a sensor, and software modules, with which only two laser-modulated images of boxes are required for volume measurement. For laser-modulated images, a novel end-to-end deep learning model is proposed by using an improved holistically nested edge detection network to extract edges. Furthermore, an automatic one-step calibration method for the line-structured light projector is designed for fast calibration. The experimental results show that the measuring range of our proposed system is 100–1800 mm, with errors less than ±5.0 mm. Theoretical analysis indicates that within the measuring range of the system, the measurement uncertainty of the measuring device is ±0.52 mm to ±4.0 mm, which is consistent with the experimental results. The device size is 140 mm × 35 mm × 35 mm and the weight is 110 g, thus the system is suitable for portable automatic box volume measurement.

As a crucial component of product quality surveillance in the industrial manufacturing field, surface defect detection plays a pivotal role in achieving industrial automation. Although numerous defect detection methods have been widely used in steel production, complicated defect characteristics still severely affect the accuracy and efficiency of defect detection methods. To solve this problem, we propose a multi-path feature aggregation network that can significantly improve steel defect detection performance. Firstly, we design an efficient multi-path aggregation feature extraction module combining gradient path with data path propagation design paradigm as the backbone feature extraction unit based on YOLOv5. Secondly, in the branch path of the backbone, deformable convolution is introduced to expand the effective receptive field and enhance the capability of extracting irregularly distributed defect features. Furthermore, we propose a chained feature enhanced pooling module that refines feature information to obtain feature maps containing more complicated defect details in the pooling layer. Experimental results on the publicly available steel surface defect datasets show that our proposed model achieves a mean average precision of 81.9 % on Northeastern University-DET and 72.1 % on GC10-DET improving by 4.6 % and 3.4 % over the baseline, respectively. The proposed defect detection network can achieve a balance between accuracy and efficiency, and meet the demand in practical production. The detection speed of our work reaches 41.1 frames per second (FPS) and 40.0 FPS on the above two datasets, which demonstrates that our method has achieved comprehensive performance in steel surface defect detection.