The early diagnosis of Alzheimer’s disease (AD) can allow patients to take preventive measures before irreversible brain damage occurs. It can be seen from cross-sectional imaging studies of AD that the features of the lesion areas in AD patients, as observed by magnetic resonance imaging (MRI), show significant variation, and these features are distributed throughout the image space. Since the convolutional layer of the general convolutional neural network (CNN) cannot satisfactorily extract long-distance correlation in the feature space, a deep residual network (ResNet) model, based on spatial transformer networks (STN) and the non-local attention mechanism, is proposed in this study for the early diagnosis of AD. In this ResNet model, a new Mish activation function is selected in the ResNet-50 backbone to replace the Relu function, STN is introduced between the input layer and the improved ResNet-50 backbone, and a non-local attention mechanism is introduced between the fourth and the fifth stages of the improved ResNet-50 backbone. This ResNet model can extract more information from the layers by deepening the network structure through deep ResNet. The introduced STN can transform the spatial information in MRI images of Alzheimer’s patients into another space and retain the key information. The introduced non-local attention mechanism can find the relationship between the lesion areas and normal areas in the feature space. This model can solve the problem of local information loss in traditional CNN and can extract the long-distance correlation in feature space. The proposed method was validated using the ADNI (Alzheimer’s disease neuroimaging initiative) experimental dataset, and compared with several models. The experimental results show that the classification accuracy of the algorithm proposed in this study can reach 97.1%, the macro precision can reach 95.5%, the macro recall can reach 95.3%, and the macro F1 value can reach 95.4%. The proposed model is more effective than other algorithms.

Influenced by glowing effects, nighttime haze removal is a challenging ill-posed task. Existing nighttime dehazing methods usually result in glowing artifacts, color shifts, overexposure, and noise amplification. Thus, through statistical and theoretical analyses, we propose a simple and effective gray haze-line prior (GHLP) to identify accurate hazy feature areas. This prior demonstrates that haze is concentrated on the haze line in the RGB color space and can be accurately projected into the gray component in the Y channel of the YUV color space. Based on this prior, we establish a new unified nighttime haze removal framework and then decompose a nighttime hazy image into color and gray components in the YUV color space. Glowing color correction and haze removal are two important consecutive steps in the nighttime dehazing process. The glowing color correction method is designed to separately remove glow in the color component and enhance illumination in the gray component. After obtaining a refined nighttime hazy image, we propose a new structure-aware variational framework to simultaneously estimate the inverted scene radiance and the transmission in the gray component. This approach can not only recover the high-quality nighttime scene radiance but also preserve the significant structural information and intrinsic color of the scene. Quantitative and qualitative comparisons validate the excellent effectiveness of the proposed nighttime dehazing method against previous state-of-the-art methods. In addition, the proposed approach can be extended to achieve image enhancement for inclement weather scenes, such as sandstorm scenes and extreme daytime hazy scenes.

Due to the atmospheric scattering and absorption, hazy weather often occurs in our everyday life, thus reducing the visibility of scenes. Single image dehazing is considered as an ill-posed and challenging problem in computer vision. To restore visibility in inclement weather, we propose an attention-to-attention generative adversarial network (AAGAN) whose motivation is the human visual perceptual mechanism. More specifically, a dense channel attention model is embedded into the encoder. Moreover, its output is projected forward to the corresponding multiscale spatial attention model in the decoder. Both attention models form an attention-to-attention mechanism to implement attention projection, thus capturing global feature dependencies of the whole network. Besides, we analyze the dehazing mechanism based on the atmospheric scattering model, and then utilize an improved RaLSGAN to recover more realistic texture information and enhance visual contrast for different hazy scenes. Finally, in order to improve the visual performance of image restoration, we remove all the instance normalization layers to avoid unnecessary artifacts, and then introduce spectral normalization for all the convolution layers to stabilize the entire training process. Qualitative assessments and analyses demonstrate that our proposed approach can achieve remarkable dehazing performance on both synthetic and real-world scenes against previous state-of-the-art methods.

High-quality and complete human motion 4D reconstruction is of great significance for immersive VR and even human operation. However, it has inevitable self-scanning constraints, and tracking under monocular settings also has strict restrictions. In this paper, we propose a human motion capture system combined with human priors and performance capture that only uses a single RGB-D sensor. To break the self-scanning constraint, we generated a complete mesh only using the front view input to initialize the geometric capture. In order to construct a correct warping field, most previous methods initialize their systems in a strict way. To maintain high fidelity while increasing the easiness of the system, we updated the model while capturing motion. Additionally, we blended in human priors in order to improve the reliability of model warping. Extensive experiments demonstrated that our method can be used more comfortably while maintaining credible geometric warping and remaining free of self-scanning constraints.

Direct estimation of body weight through noncontact methods is crucial for applications such as health monitoring, surveillance, and robot-assisted casualty rescue. Existing methods for body weight estimation from images are often affected by various factors, such as camera distance, human orientation, and body posture, etc., which ignore the fact that bodies typically inhabit 3D space. To address the problem, we propose an approach for estimating body weight based on virtual anthropometric measurements and deep features instead of estimating by reasoning pixels. Specifically, we develop a three-branch framework that includes face feature extraction, body feature extraction as well as deep feature extraction, and maps all features with a regressor. The developed method adds 3D shape reconstruction to explicitly reason about virtual anthropometric measurements. To enable this, our model is trained to robustly compute anthropometric measurements in various orientations and postures. Furthermore, we evaluate our method on a public dataset and Image-VM-BMI, a new dataset of 4740 images, including BMI labels and virtual anthropometric measurement labels with paired 3D reconstruction. Extensive experimental results demonstrate that the proposed method outperforms pixel-based analysis approaches on BMI estimation.

When disasters such as earthquakes and collapses occur, rescuers must go deep into dangerous regions and rush to transport the wounded. Long-term rescue operations in dangerous environments challenge the rescuers’ physical strength and threaten their lives. We propose a human-body reconstruction method for autonomous rescue operations with a wounded rescue robot. This method constructs a human body model that can support a robot’s autonomous rescue operations given limited scans of the injured. Most casualties lie on the ground in supine positions, so this scanning limit is typical. We propose a scanning viewpoint planning method using deep reinforcement learning and change the viewpoint with the help of a 6-degree-of-freedom (DOF) robotic arm to obtain a complete scan of the visible surface. For successful autonomous casualty rescue, a robot must have access to a watertight human body mesh with precise surfaces and associated semantics. To this end, we develop a neural network that predicts a completely clothed body mesh from visible point clouds. The results of experiments show that our method can reconstruct a watertight body mesh with accurate surfaces and human semantics to satisfy the needs of autonomous robotic casualty rescue operations under scanning limitations.