Multi-class GAN for generating multi-class images in object recognition.

The current generative adversarial network (GAN) is limited in the application of data augmentation in object recognition. The training of the GAN is unstable, and the generated image quality is poor. Methods such as progressive growing of GANs and multi-scale gradient GAN solve these problems. The packed GAN (PacGAN) solves the problem of mode collapse during training. However, these methods can generate only one type of image at a time, and the training time is long. To solve the above problems, this paper proposes the multi-class GAN (Mc-GAN). It uses an augmented discriminator to train multiple generators at the same time. Through iterative training, the discriminator can accurately judge the output of each generator and guide it to generate the corresponding image. This paper analyzes the optimization process of the objective function of Mc-GAN. Experiments show that the method can generate high-quality images and reduce training time, and it can be used for data augmentation in object recognition. It effectively improves the practicality of GAN.

Robust muscle force prediction using NMFSEMD denoising and FOS identification.

In this paper, an aliasing noise restraint technique and a system identification-based surface electromyography (sEMG)-force prediction model are proposed to realize a type of robust sEMG and muscle force prediction. For signal denoising, a novel non-negative matrix factorization screening empirical mode decomposition (NMFSEMD) and a fast orthogonal search (FOS)-based muscle force prediction model are developed. First, the NMFSEMD model is used to screen the empirical mode decomposition (EMD) results into the noisy intrinsic mode functions (IMF). Then, the noise matrix is computed using IMF translation and superposition, and the matrix is used as the input of NMF to obtain the denoised IMF. Furthermore, the reconstruction outcome of the NMFSEMD method can be used to estimate the denoised sEMG. Finally, a new sEMG muscle force prediction model, which considers a kind of candidate function in derivative form, is constructed, and a data-training-based linear weighted model is obtained. Extensive experimental results validate the suggested method's correction: after the NMFSEMD denoising of raw sEMG signal, the signal-noise ratio (SNR) can be improved by about 15.0 dB, and the energy percentage (EP) can be greater than 90.0%. Comparing with the muscle force prediction models using the traditional pretreatment and LSSVM, and the NMFSEMD plus LSSVM-based method, the mean square error (MSE) of our approach can be reduced by at least 1.2%.

JRL-YOLO: A Novel Jump-Join Repetitious Learning Structure for Real-Time Dangerous Object Detection.

Campus security incidents occur from time to time, which seriously affect the public security. In recent years, the rapid development of artificial intelligence has brought technical support for campus intelligent security. In order to quickly recognize and locate dangerous targets on campus, an improved YOLOv3-Tiny model is proposed for dangerous target detection. Since the biggest advantage of this model is that it can achieve higher precision with very fewer parameters than YOLOv3-Tiny, it is one of the Tinier-YOLO models. In this paper, the dangerous targets include dangerous objects and dangerous actions. The main contributions of this work include the following: firstly, the detection of dangerous objects and dangerous actions is integrated into one model, and the model can achieve higher accuracy with fewer parameters. Secondly, to solve the problem of insufficient YOLOv3-Tiny target detection, a jump-join repetitious learning (JRL) structure is proposed, combined with the spatial pyramid pooling (SPP), which serves as the new backbone network of YOLOv3-Tiny and can accelerate the speed of feature extraction while integrating features of different scales. Finally, the soft-NMS and DIoU-NMS algorithm are combined to effectively reduce the missing detection when two targets are too close. Experimental tests on self-made datasets of dangerous targets show that the average MAP value of the JRL-YOLO algorithm is 85.03%, which increases by 3.22 percent compared with YOLOv3-Tiny. On the VOC2007 dataset, the proposed method has a 9.29 percent increase in detection accuracy compared to that using YOLOv3-Tiny and a 2.38 percent increase compared to that employing YOLOv4-Tiny, respectively. These results all evidence the great improvement in detection accuracy brought by the proposed method. Moreover, when testing the dataset of dangerous targets, the model size of JRL-YOLO is 5.84 M, which is about one-fifth of the size of YOLOv3-Tiny (33.1 M) and one-third of the size of YOLOv4-Tiny (22.4 M), separately.

Spectral Unmixing of Hyperspectral Remote Sensing Imagery via Preserving the Intrinsic Structure Invariant.

Hyperspectral unmixing, which decomposes mixed pixels into endmembers and corresponding abundance maps of endmembers, has obtained much attention in recent decades. Most spectral unmixing algorithms based on non-negative matrix factorization (NMF) do not explore the intrinsic manifold structure of hyperspectral data space. Studies have proven image data is smooth along the intrinsic manifold structure. Thus, this paper explores the intrinsic manifold structure of hyperspectral data space and introduces manifold learning into NMF for spectral unmixing. Firstly, a novel projection equation is employed to model the intrinsic structure of hyperspectral image preserving spectral information and spatial information of hyperspectral image. Then, a graph regularizer which establishes a close link between hyperspectral image and abundance matrix is introduced in the proposed method to keep intrinsic structure invariant in spectral unmixing. In this way, decomposed abundance matrix is able to preserve the true abundance intrinsic structure, which leads to a more desired spectral unmixing performance. At last, the experimental results including the spectral angle distance and the root mean square error on synthetic and real hyperspectral data prove the superiority of the proposed method over the previous methods.

Restoration of motion-blurred image based on border deformation detection: a traffic sign restoration model.

Due to the rapid development of motor vehicle Driver Assistance Systems (DAS), the safety problems associated with automatic driving have become a hot issue in Intelligent Transportation. The traffic sign is one of the most important tools used to reinforce traffic rules. However, traffic sign image degradation based on computer vision is unavoidable during the vehicle movement process. In order to quickly and accurately recognize traffic signs in motion-blurred images in DAS, a new image restoration algorithm based on border deformation detection in the spatial domain is proposed in this paper. The border of a traffic sign is extracted using color information, and then the width of the border is measured in all directions. According to the width measured and the corresponding direction, both the motion direction and scale of the image can be confirmed, and this information can be used to restore the motion-blurred image. Finally, a gray mean grads (GMG) ratio is presented to evaluate the image restoration quality. Compared to the traditional restoration approach which is based on the blind deconvolution method and Lucy-Richardson method, our method can greatly restore motion blurred images and improve the correct recognition rate. Our experiments show that the proposed method is able to restore traffic sign information accurately and efficiently.

A novel abandoned object detection system based on three-dimensional image information.

A new idea of an abandoned object detection system for road traffic surveillance systems based on three-dimensional image information is proposed in this paper to prevent traffic accidents. A novel Binocular Information Reconstruction and Recognition (BIRR) algorithm is presented to implement the new idea. As initial detection, suspected abandoned objects are detected by the proposed static foreground region segmentation algorithm based on surveillance video from a monocular camera. After detection of suspected abandoned objects, three-dimensional (3D) information of the suspected abandoned object is reconstructed by the proposed theory about 3D object information reconstruction with images from a binocular camera. To determine whether the detected object is hazardous to normal road traffic, road plane equation and height of suspected-abandoned object are calculated based on the three-dimensional information. Experimental results show that this system implements fast detection of abandoned objects and this abandoned object system can be used for road traffic monitoring and public area surveillance.

A novel multisensor traffic state assessment system based on incomplete data.

A novel multisensor system with incomplete data is presented for traffic state assessment. The system comprises probe vehicle detection sensors, fixed detection sensors, and traffic state assessment algorithm. First of all, the validity checking of the traffic flow data is taken as preprocessing of this method. And then a new method based on the history data information is proposed to fuse and recover the incomplete data. According to the characteristics of space complementary of data based on the probe vehicle detector and fixed detector, a fusion model of space matching is presented to estimate the mean travel speed of the road. Finally, the traffic flow data include flow, speed and, occupancy rate, which are detected between Beijing Deshengmen bridge and Drum Tower bridge, are fused to assess the traffic state of the road by using the fusion decision model of rough sets and cloud. The accuracy of experiment result can reach more than 98%, and the result is in accordance with the actual road traffic state. This system is effective to assess traffic state, and it is suitable for the urban intelligent transportation system.

Spatial-resolution analysis and optimal design of integral imaging.

Integral imaging is a promising technology for 3D imaging and display. This paper reports the 3D spatial-resolution research based on reconstructed 3D space. Through geometric analysis of the reconstructed optical distribution from all the element images that attend recording, the relationship among microlens parameters, planar-recording resolution, and 3D spatial resolution was obtained. The effect of microlens parameter accuracy on the reconstructed position error also was discussed. The research was carried on the depth priority integral imaging system (DPII). The results can be used in the optimal design of integral imaging.