Search

This paper introduces a related model of membrane calculation in the defect detection and positioning of industrial components. It has the characteristics of distributed and parallel computing, and can efficiently search for better solutions in a given feature space. Inspired by the membrane clustering algorithm, this paper proposes a greedy membrane clustering algorithm and names it GMCA. GMCA is applied after the extraction of local features of normal samples. It uses a greedy strategy to construct a sub-feature set that describes the local characteristics of normal samples. During training, GMCA can learn the membrane cluster center of normal image blocks and each sub-feature within the cluster. At test time, the anomaly map is obtained by calculating the distance from the test sample block to the corresponding cluster center and the maximum distance from the cluster center to the nearest neighbor in the training sample. This solves the limitation of traditional algorithms requiring dataset alignment. In the unsupervised dataset MvTec AD, samples can be divided into object categories and texture categories according to the background of images. The pixel-level anomaly location index (AUROC) of this method on object category data reaches 98.3%. The image-level anomaly detection index (AUROC) on texture category data reaches 99.1%. • We design a computational model of membrane clustering using the evolutionary mechanisms and communicative mechanisms of cells. • GMCA has the global optimization characteristics of high accuracy and fast convergence of the membrane clustering algorithm. • GMCA has the local optimization characteristics of the greedy strategy. • GMCA solves the limitation of traditional defect detection and positioning methods that require dataset alignment. • Numerous experiments show the proposed GMCA performs competitively in industrial defect detection and location prediction. [ABSTRACT FROM AUTHOR]

Currently, image segmentation is widely used in face recognition, medical imaging, traffic control systems, and many other fields. The traditional possibilistic C-means (PCM) algorithm reduces the impact of outliers and noise on the computation of clustering centers; however, the clustering results are still poor due to high noise points. In this paper, an interval possibilistic C-means (IVPCM) algorithm is proposed that expands the natural number of image pixels to an interval value. In addition, a method to calculate the sample distance is proposed. Then, secondary feature extraction of interval values is conducted to improve the clustering. Several experiments show that IVPCM has more advantages in segmentation performance. [ABSTRACT FROM AUTHOR]

A method to enhance Web service clustering is proposed in this paper. Since current service clustering methods usually face low quality of service representation vectors and lack consideration of service collaboration, we try to provide an improved topic model to generate high-quality service representation vectors and design a service clustering method to integrate function similarity and collaboration similarity. First, by introducing feature word extraction and probability distribution correction into GSDMM, we present a model called TE-GSDMM (topic enhanced Gibbs sampling algorithm for the Dirichlet Multinomial Mixture model). Then, a service collaboration graph is put forward to model cooperation relationships and generate service collaboration vectors. Collaboration similarity is assessed by the similarity of service collaboration vectors. Finally, the K-means++ algorithm is employed to cluster Web services by evaluating service function similarity and collaboration similarity. Experiments show that TE-GSDMM outperforms other topic models in generating high-quality service representation vectors for service clustering. Moreover, service clustering performance is further improved by integrating collaboration similarity. Thus, the proposed method effectively enhances Web service clustering by improving the quality of service representation vectors and integrating service collaboration similarity. [ABSTRACT FROM AUTHOR]

It is widely accepted that webpage layout is important for website aesthetics. However, knowledge is limited about which kind of webpage layout is aesthetic. In this paper, we conduct a large-scale analysis of 13,017 real webpages with different aesthetic levels. The results demonstrate that there are significant differences between aesthetic webpages and non-aesthetic webpages in several aspects of layout design, such as the position of objects and the average number of objects. In order to identify whether the web layout is aesthetic, we propose a novel Webpage Layout Aesthetic Evaluation model (WLAE) based on an improved Adaboost algorithm to automatically predict the aesthetics of a webpage layout. Experimental results show that our WLAE model is significantly better than other existing methods. Based on the feature analysis of 13,017 real-world webpages and the contribution of each feature to webpage aesthetic prediction, we have summarized 4 webpage design principles for webpage designers. Besides, a large number of real webpages collected by our team can be used for further research on webpage layout aesthetic design and/or related research. [ABSTRACT FROM AUTHOR]

Recently, robust-norm distance related principal component analysis (PCA) for feature extraction has been shown to be very effective for image analysis, which considers either minimization of reconstruction error or maximization of data variance in low-dimensional subspace. However, both of them are important for feature extraction. Furthermore, most of existing methods cannot obtain satisfactory results due to the utilization of inflexible robust norm for distance metric. To address these problems, this paper proposes a novel robust PCA formulation called Double L 2,p -norm based PCA (DLPCA) for feature extraction, in which the minimization of reconstruction error and the maximization of variance are simultaneously taken into account in a unified framework. In the reconstruction error function, we target to learn a latent subspace to bridge the relationship between the transformed features and the original features. To guarantee the objective to be insensitive to outliers, we take L 2,p -norm as the distance metric for both reconstruction error and data variance. These characteristics make our method more applicable for feature extraction. We present an effective iterative algorithm to obtain the solution of this challenging work, and conduct theoretical analysis on the convergence of the algorithm. The experimental results on several databases show the effectiveness of our model. [ABSTRACT FROM AUTHOR]

Stereo vision is widely deployed on robots and drones to enable depth estimation at a low cost. The combination of lightweight deep neural network (DNN) and cost volumes algorithm is proved to possess the advantages of both high depth estimation accuracy and speed. However, currently there is no accelerator architecture compatible with both efficient DNN inference and cost generation algorithms such as stereo matching. This work proposes a stereo vision accelerator called Dadu-eye, dedicated to real-time processing of high-resolution image streams. The proposed architecture adopts a pipelined hardware design with the techniques of operation approximation and scheduling-level optimization. First, a cost estimation block is designed to generate cost volumes from both luminance and color information. Second, a super pipelined multiplication and accumulation array with a row scan-based fused-layer convolution scheduling is proposed to perform the encoding and decoding neural network efficiently. Finally, an optical flow block is designed and cooperates with the array to approximately predict half of the frames’ depth to achieve real-time (30fps) processing on 1080p view. Based on the SMIC 40 nm CMOS process, this stereo vision accelerator achieves 5.3 TOPS/W power efficiency and significantly reduces 81% off-chip memory access. [ABSTRACT FROM AUTHOR]