International Journal of Pattern Recognition and Artificial Intelligence

A multi-feature broad learning system (MFBLS) is proposed to improve the image classification performance of broad learning system (BLS) and its variants. The model is characterized by two major characteristics: multi-feature extraction method and parallel structure. Multi-feature extraction method is utilized to improve the feature-learning ability of BLS. The method extracts four features of the input image, namely convolutional feature, K-means feature, HOG feature and color feature. Besides, a parallel architecture that is suitable for multi-feature extraction is proposed for MFBLS. There are four feature blocks and one fusion block in this structure. The extracted features are used directly as the feature nodes in the feature block. In addition, a “stacking with ridge regression” strategy is applied to the fusion block to get the final output of MFBLS. Experimental results show that MFBLS achieves the accuracies of 92.25%, 81.03%, and 54.66% on SVHN, CIFAR-10, and CIFAR-100, respectively, which outperforms BLS and its variants. Besides, it is even superior to the deep network, convolutional deep belief network, in both accuracy and training time on CIFAR-10. Code for the paper is available at https://github.com/threedteam/mfbls.

Fatigue driving is one of the main causes of traffic accidents. In recent years, considerable attention has been paid to fatigue detection systems, which is an important solution for preventing fatigue driving. In order to prevent and reduce fatigue driving, a driver fatigue detection system based on computer vision is proposed. In this system, an improved face detection method is used to detect the driver’s face from the image obtained by a charge coupled device (CCD) camera. Then, the feature points of the eyes and mouth are located by an ensemble of regression trees. Next, fatigue characteristic parameters are calculated by the improved percentage of eyelid closure over the pupil over time algorithm. Finally, the state of drivers is evaluated by using a fuzzy neural network. The system can effectively monitor and remind the state of drivers so as to significantly avoid or decrease the occurrence of traffic accidents. The experimental results show that the system is of wonderful real-time performance and accurate recognition rate, so it meets the requirements of practicality in driver fatigue detection greatly.

Unlabeled samples and transformation matrix are two main parts of unsupervised and semi-supervised feature extraction (FE) algorithms. In this manuscript, a semi-supervised FE method, locality preserving projection in the probabilistic framework (LPPPF), to find a sufficient number of reliable and unmixed unlabeled samples from all classes and constructing an optimal projection matrix is proposed. The LPPPF has two main steps. In the first step, a number of reliable unlabeled samples are selected based on the training samples, spectral features, and spatial information in the probabilistic framework. In this way, the spectral and spatial probability distribution function is calculated for each unlabeled sample. Therefore, the spectral features and spatial information are integrated together with a joint probability distribution function. Finally, a sufficient number of unlabeled samples with the highest joint probability distribution are selected. In the second step, the selected unlabeled samples are applied to construct the transformation matrix based on the spectral and spatial information of the unlabeled samples. The adjacency graph is improved by using new weights based on spectral and spatial information. This method is evaluated on three data sets: Indian Pines, Pavia University, and Kennedy Space Center (KSC) and compared with some recent and well-known supervised, semi-supervised, and unsupervised FE methods. Various experiments demonstrate the efficiency of the LPPPF in comparison with the other FE methods. LPPPF has also considerable performance with limited training samples.

As a decisive parameter of network robustness and network economy, the capacity of network edges can directly affect the operation stability and the construction cost of the network. This paper proposes a multilevel load–capacity optimal relationship (MLCOR) model that can substantially improve the network economy on the premise of network safety. The model is verified in artificially created networks including free-scale networks, small-world networks, and in the real network structure of the Shanghai Metro network as well. By numerical simulation, it is revealed that under the premise of ensuring the stability of the network from the destruction caused by initial internal or external damage on edge, the MLCOR model can effectively reduce the cost of the entire network compared to the other two linear load–capacity models regardless of what extent of the destruction that the network edges suffer initially. It is also proved that there exists an optimal tunable parameter and the corresponding optimal network cost for any BA and NW network topology, which can provide the reference for setting reasonable capacities for network edges in a real network at the stage of network planning and construction, promoting security and stability of network operation.

Similarity/distance measures play a key role in many machine learning, pattern recognition, and data mining algorithms, which leads to the emergence of the metric learning field. Many metric learning algorithms learn a global distance function from data that satisfies the constraints of the problem. However, in many real-world datasets, where the discrimination power of features varies in the different regions of input space, a global metric is often unable to capture the complexity of the task. To address this challenge, local metric learning methods are proposed which learn multiple metrics across the different regions of the input space. Some advantages of these methods include high flexibility and learning a nonlinear mapping, but they typically achieve at the expense of higher time requirements and overfitting problems. To overcome these challenges, this research presents an online multiple metric learning framework. Each metric in the proposed framework is composed of a global and a local component learned simultaneously. Adding a global component to a local metric efficiently reduces the problem of overfitting. The proposed framework is also scalable with both sample size and the dimension of input data. To the best of our knowledge, this is the first local online similarity/distance learning framework based on Passive/Aggressive (PA). In addition, for scalability with the dimension of input data, Dual Random Projection (DRP) is extended for local online learning in the present work. It enables our methods to run efficiently on high-dimensional datasets while maintaining their predictive performance. The proposed framework provides a straightforward local extension to any global online similarity/distance learning algorithm based on PA. Experimental results on some challenging datasets from machine vision community confirm that the extended methods considerably enhance the performance of the related global ones without increasing the time complexity.

Compressive Sensing for Magnetic Resonance Imaging (CS-MRI) aims to reconstruct Magnetic Resonance (MR) images from under-sampled raw data. There are two challenges to improve CS-MRI methods, i.e. designing an under-sampling algorithm to achieve optimal sampling, as well as designing fast and small deep neural networks to obtain reconstructed MR images with superior quality. To improve the reconstruction quality of MR images, we propose a novel deep convolutional neural network architecture for CS-MRI named MRCSNet. The MRCSNet consists of three sub-networks, a compressive sensing sampling sub-network, an initial reconstruction sub-network, and a refined reconstruction sub-network. Experimental results demonstrate that MRCSNet generates high-quality reconstructed MR images at various under-sampling ratios, and also meets the requirements of real-time CS-MRI applications. Compared to state-of-the-art CS-MRI approaches, MRCSNet offers a significant improvement in reconstruction accuracies, such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM). Besides, it reduces the reconstruction error evaluated by the Normalized Root-Mean-Square Error (NRMSE). The source codes are available at https://github.com/TaihuLight/MRCSNet.

Image scene analysis is to analyze image scene content through image semantic segmentation, which can identify the categories and positions of different objects in an image. However, due to the loss of spatial detail information, the accuracy of image scene analysis is often affected, resulting in rough edges of FCN, inconsistent class labels of target regions and missing small targets. To address these problems, this paper increases the receptive field, conducts multi-scale fusion and changes the weight of different sensitive channels, so as to improve the feature discrimination and maintain or restore spatial detail information. Furthermore, the deep neural network FCN is used to build the base model of semantic segmentation. The ASPP, data augmentation, SENet, decoder and global pooling are added to the baseline to optimize the model structure and improve the effect of semantic segmentation. Finally, the more accurate results of scene analysis are obtained.

This paper proposes a research on cross-media intelligent perception and retrieval analysis technology based on deep learning education in view of the complex learning knowledge environment of artificial intelligence of traditional cross-media retrieval technology, which is unable to obtain retrieval information timely and accurately. Based on the cross-media theory, this paper analyzes the pre-processing of cross-media data, transforms from a single media form such as text, voice, image, and video to a cross-media integration covering network space and physical space, and designs a cross-media intelligent perception platform system. With multi-core method of typical correlation analysis algorithm, this paper develops a new joint learning framework based on the joint feature selection and subspace learning cross-modal retrieval. What’s more, the algorithm is tested experimentally. The results show that the retrieval analysis technology is significantly better than the traditional media retrieval technology, which can effectively identify text semantics and visual image classification, and can better maintains the relevance of data content and the consistency of semantic information, which has important reference value for cross-media applications.

Establishment of a performance evaluation model is a hotspot of current research. In this paper, the performance bottleneck is analyzed quantitatively, which provided programmers with a guidance to optimize the performance bottleneck. This paper takes a matrix as an example; the matrix is divided into a dense matrix or a sparse matrix. For dense matrix, the performance is first analyzed in a quantitative way, and an evaluation model is developed, which includes the instruction pipeline, shared memory, and global memory. For sparse matrix, this paper aims at the four formats of CSR, ELL, COO, and HYB, through the observation data obtained from the actual operation of large datasets, finds the relationship between the running time, dataset form, and storage model, and establishes their relational model functions. Through practical test and comparison, the error between the execution time of the test dataset that is predicted by the model function and the actual running time is found to be within a stable finite deviation threshold, proving that the model has certain practicability.

This paper designs a method that can generate anchors of various shapes for the object detection framework. This method has the characteristics of novelty and flexibility. Different from the previous anchors generated by a pre-defined manner, our anchors are generated dynamically by an anchor generator. Specially, the anchor generator is not fixed but learned from the hand-designed anchors, which means that our anchor generator is able to work well in various scenes. In the inference time, the weights of anchor generator are estimated by a simple network where the input is some hand-designed anchor. In addition, in order to make the difference between the number of positive and negative samples smaller, we use an adaptive IOU threshold related to the object size to solve this problem. At the same time, we proved that our proposed method is effective and conducted a lot of experiments on the COCO dataset. Experimental results show that after replacing the anchor generation method in the previous object detectors (such as SSD, mask RCNN, and Retinanet) with our proposed method, the detection performance of the model has been greatly improved compared to before the replacement, which proves our method is effective.

Traffic congestion has become a major problem restricting the development of major cities. ITS (Intelligent Transportation System) can record the state of traffic and predict the future traffic state, then reasonably optimize the travel scheme, so as to achieve the purpose of alleviating traffic congestion. Meanwhile, traffic flow prediction can provide data support for ITS, so many researchers have done a lot of research on traffic flow prediction. Many researchers take the traffic network as an undirected graph, and use the GCN (Graph Convolution Network) model to study the traffic flow prediction, and have achieved good prediction results. However, the traffic network is directed, and the traffic network is regarded as an undirected graph, which loses the direction information of the road network. Therefore, this inspires us to propose a graph convolution operator DGCN (Directed GCN), which can make full use of the in degree and out degree information of each station in the traffic network. The experimental results show that the graph convolution neural network based on this operator has better prediction accuracy than the state-of-the-art models.

This study aims to explore the influence of various possible combinations of consumer attribute benefits and the message direction provided by merchants on the final purchase intention of consumers. A two-factor experimental design was used to collect data from Generation Z consumers. Four experimental combination scenarios were designed based on consumer attribute benefits (hedonic or utilitarian) and message direction (positive or negative messages). A total of 160 valid samples were obtained, and the influence of variable-final purchase intention was analyzed. The results showed that the combination of product attributes and message direction has a significant impact on purchase intention and that hedonic products should be recommended to process-benefit consumers through positive messages. However, chatbots should choose utilitarian products and negative messages to effectively improve consumers’ purchase intention and willingness. In practical application, by finding out the best correspondence method of message transmission and providing the AI chatbot with a reference for the logical construction of content transmission, the communication efficiency between merchants and consumers can be enhanced, and the scientific and technological service capability of merchants can be improved.