Research on Multiscale Pedestrian Detection Algorithm

Infants and children with respiratory conditions are often prescribed bronchodilators. Face masks are used to facilitate the administration of nebulized therapy in patients unable to use a mouthpiece. Masks incorporate holes into their design, and their occlusion during aerosol delivery has been a common practice. Masks are available in different sizes and different dead volumes. The aim of this study was to compare the effect of different degrees of occlusion of the mask holes and different mask dead space on the amount of nebulized albuterol available at the mouth opening in a model of a spontaneously breathing child. A breathing simulator mimicking infant (tidal volume [VT] = 50 mL, breathing frequency = 30 breaths/min, inspiratory-expiratory ratio [I:E] = 1:3), child (VT = 155 mL, breathing frequency = 25 breaths/min, I:E = 1:2), and adult (VT = 500 mL, breathing frequency = 15 breaths/min, I:E = 1:2) breathing patterns was connected to a collection filter hidden behind a face plate. A pediatric size mask and an adult size mask connected to a continuous output jet nebulizer were sealed to the face plate. Three nebulizers were loaded with albuterol sulfate (2.5 mg/3 mL) and operated with 6 L/min compressed air for 5 min. Experiments were repeated with different degrees of occlusion (0%, 50%, and 90%). Albuterol was extracted from the filter and measured with a spectrophotometer at 276 nm. Occlusion of the holes in the large mask did not increase the amount of albuterol in any of the breathing patterns. The amount of albuterol captured at the mouth opening did not change when the small mask was switched to the large mask, except with the breathing pattern of a child, and when the holes in the mask were 50% occluded (P = .02). Neither decreasing the dead space of the mask nor occluding the mask holes increased the amount of nebulized albuterol captured at the mouth opening.

With the rapid development of mobile communication technology, ranging and angling methods based on LTE and 5G signals have been widely used in location-based services (LBSs). However, mobile communication signals also undergo different degrees of occlusion and refraction during propagation because of the influence of buildings and human activities. The conventional ranging and angling methods are unable to extract high-precision navigation observations from these non-line-of-sight (NLOS) signals. Regarding the NLOS signals that seriously affect the performance of wireless positioning systems, this paper proposes a deep learning method based on the self-attention mechanism by distinguishing the difference characteristics in the autocorrelation functions of line-of-sight (LOS) and NLOS signals. It provides a basis for developing a stable LBS by accurately determining the state of received signals in complex indoor and outdoor environments. The simulation and field tests were carried out separately to evaluate the performance of the method developed in this paper under different scenarios, which achieved a discrimination accuracy better than 99.9% for NLOS signals.

Observational ergonomic postural assessment methods have been used to evaluate the risks of musculoskeletal disorders. Recently, researchers have actively proposed semiautomatic approaches using motion capture data. This study compared joint angle measurements from optical, inertial, and depth-based motion capture systems, especially for cases with occlusions. Participants performed 6 static postures with different degrees of occlusion while capturing motion data by three motion capture systems. Results showed that the joint angle difference between the three systems was acceptably low in general, but the difference between Kinect and the other two systems was larger, especially in cases with occlusions. The findings indicate that the Kinect-based system is less stable than optical and inertial-based motion capture systems, but it can be used for ergonomic postural assessments in the environment without severe occlusions.

At present there are some worm detection systems, primarily for a single LAN or with hardware router environment, which main use worm propagation characteristics for detection and has high false alarm rate, but it is not applicable for large-scale network for detecting. This paper presents a distributed worm detection technology, which is divided into two parts, client end and the console end program. The system uses the rule-based detection methods to monitor network worms, and the console side manages and coordinates the multiple side work of detecting. Experimental results show that the method can be good applicable for worms conduct surveillance based on a single or multiple local area network and is used for worm alarming, the method has high detection rate and low false alarm rate.

At present there are some worm intrusion detection systems, primarily for a single LAN or with hardware router environment, which are not applicable for large-scale network detection or have high false alarm rate by using only worm propagation characteristics for detection. This paper analyzed worm non-linear propagation models and drew out the worm transmission curves. Then a distributed worm detection technology is designed. The novel distributed worm detection system consists of two parts, client end and console end programs. The system uses rule-based detection method to monitor network worms, and the console side manages and coordinates detection work of the client sides. Experimental results show that the technology is a good solution to worm detection in multiple network environments which can give an alarm with high detection rate and low false alarm rate when the known worm appears.

The target tracking by space-based surveillance systems is difficult due to the long distances, weak energies, fast speeds, high false alarm rates, and low algorithmic efficiencies involved in the process. To mitigate the impact of these difficulties, this article proposes a target tracking algorithm based on image processing and Transformer, which employs a two-dimensional Gaussian soft-thresholding method to reduce the image noise, and combines a Laplace operator-weighted fusion method to augment the image, so as to improve the overall quality of the image and increase the accuracy of target tracking. Based on the SiamCAR framework, the Transformer model in the field of natural language processing is introduced, which can be used to enhance the image features extracted from the backbone network by mining the rich temporal information between the initial and dynamic templates. In order to capture the information of the target’s appearance change in the temporal sequence, a template update branch is introduced at the input of the algorithm, which realizes the dynamic update of the templates by constructing a template memory pool, and selecting the best templates for the candidate templates in the memory pool using the cosine similarity-based selection, thus ensuring the robustness of the tracking algorithm. The experimental results that compared with the SiamCAR algorithm and the mainstream algorithms, the TrD-Siam algorithm proposed in this article effectively improves the tracking success rate and accuracy, addressing poor target tracking performance under space-based conditions, and has a good value of application in the field of optoelectronic detection.

While some image textures can be changed with scale, others cannot. We focus on a multi-scale features of determing the sensitivity of the texture intensity to change. This paper presents a new method of texture structure classification and depth estimation using multi-scale features extracted from a higher order of the local autocorrelation functions. Multi-scale features consist of the meansand variances of distributions, which are extracted from theautocorrelation feature vectors according to multi-level scale. In order to reduce dimensional feature vectors, we employ the Principal Component Analysis (PCA) in the autocorrelation feature space. Each training image texture makes its own multi-scale model in a reduced PCA feature space, and the test of the texture image is projected in the homogeneous PCA space of the training data. The experimental results show that the proposed multi-scale feature can be utilized notonly for texture classification, but also depth estimation in two dimensional images with texture gradients.

The fully convolutional networks (FCNs) have been widely applied in numerous medical image segmentation tasks. However, tissue regions usually have large variations of shape and scale, so the ability of neural networks to learn multi-scale features is important to the segmentation performance. In this paper, we improve the network for multi-scale feature fusion, in the medical image segmentation by introducing two feature fusion modules: i) global attention multi-scale feature fusion module (GMF); ii) local dense multi-scale feature fusion module (LMF). GMF aims to use global context information to guide the recalibration of low-level features from both spatial and channel aspects, so as to enhance the utilization of effective multi-scale features and suppress the noise of low-level features. LMF adopts bottom-up top-down structure to capture context information, to generate semantic features, and to fuse feature information at different scales. LMF can integrate local dense multi-scale context features layer by layer in the network, thus improving the ability of network to encode interdependent relationships among boundary pixels. Based on the above two modules, we propose a novel medical image segmentation framework (GLF-Net). We evaluated the proposed network and modules on challenging brain tumor segmentation and pancreas segmentation datasets, and very competitive performance has been achieved.

Most of existing local-based person re-identification (re-ID) methods extract powerful feature representations from multiple body regions of a pedestrian. However, local-based person re-ID methods employ horizontal or grid slicing to divide pedestrian images, which can easily cause network redundancy and misalignment of pedestrian body parts. Attention mechanisms have been recently introduced to computer vision tasks to strengthen local salient features and suppress irrelevant features simultaneously. However, attention mechanisms overemphasize local salient features, which may lead to potential significant information being missed. To address the above issues, we proposed a local information interaction enhancement network (LIEN) for person re-ID. The network integrates attention mechanism with local-based strategy to achieve mutually facilitating and complementary effects. Specifically, LIEN was composed of three parts: a backbone network, a dual attention module (DAM), and a local information interaction module (LIM). DAM was used to guide the backbone network to extract the local salient information, and LIM was used to mine the potential significant information that DAM ignored by exploring the correlations between different channel-level local features. With the combination of DAM and LIM, LIEN mined the salient features while obtaining more potential effective features. In particular, the way that it extracts local features, LIEN effectively avoided pedestrian misalignment caused by uniformly dividing pedestrian images. We evaluated the performance of LIEN on public person datasets, and experimental results indicated that LIEN achieved a more advanced performance.

Handwritten mathematical expression recognition (HMER) is a challenging task due to the complex two-dimensional structure of mathematical expressions and the similarity of handwritten texts. Most existing methods for HMER only consider single-scale features while ignoring multi-scale features that are very important to HMER. Few works have explored the fusion of multi-scale features in HMER, but exhibited an extra branch that brings more parameters and computation. In this paper, we propose an end-to-end method to integrate multi-scale features using a unified model. Specifically, we customized the Dense Atrous Spatial Pyramid Pooling (DenseASPP) to our backbone network to capture the multi-scale features of the input image meanwhile expanding the receptive fields. Moreover, we added a symbol classifier using focal loss to better discriminate and recognize similar symbols, to further improve the performance of HMER. Experiments on the Competition on Recognition of Online Handwritten Mathematical Expressions (CROHME) 2014, 2016 and 2019 shows that the proposed method achieves superior performance to most state-of-the-art methods, demonstrating the effectiveness of the proposed method.

Effectively capturing the multi-scale, multi-channel, and boundary features of target images is beneficial for improving the accuracy of salient object detection. A multichannel feature enhancement module (MFEM) is constructed by concatenating two feature maps along the channel, performing convolution activation, dot product addition, and convolution operations. An edge feature enhancement module (EFEM) is constructed by performing an inverse attention mechanism and Gaussian-Laplace transform on the predicted map, channel-wise concatenation, and convolution operations with the feature submap. Multi-scale features are extracted using a SwinTransformer encoder and enhanced using the MFEM. The enhanced features are then fused using a cross-attention mechanism for multi-level features. The fused feature map, the high-frequency feature map obtained by Gaussian-Laplace transforming the predicted map, and the external edge features are input into the EFEM to obtain the salient object detection results. This results in a novel salient object detection model with multi-scale, multi-channel, and boundary feature enhancement: MfENet. Experimental results show that the Mean Absolute Error (MAE), S-measure and F-measure accuracies of our method on the HKU-IS dataset are 0.018,0.924 and 0.959 respectively. These results demonstrate the effectiveness of our method. Furthermore, this work highlights the importance of multi-scale and boundary awareness in enhancing salient object detection, and its contribution to the technical field relevant to the theme of this conference.

A multi-scale semantic feature fusion method for remote sensing crop classification

Efficient feature representation is the key to improving crowd counting performance. CNN and Transformer are the two commonly used feature extraction frameworks in the field of crowd counting. CNN excels at hierarchically extracting local features to obtain a multi-scale feature representation of the image, but it struggles with capturing global features. Transformer, on the other hand, could capture global feature representation by utilizing cascaded self-attention to capture remote dependency relationships, but it often overlooks local detail information. Therefore, relying solely on CNN or Transformer for crowd counting has certain limitations. In this paper, we propose the TCHNet crowd counting model by combining the CNN and Transformer frameworks. The model employs the CMT (CNNs Meet Vision Transformers) backbone network as the Feature Extraction Module (FEM) to hierarchically extract local and global features of the crowd using a combination of convolution and self-attention mechanisms. To obtain more comprehensive spatial local information, an improved Progressive Multi-scale Learning Process (PMLP) is introduced into the FEM, guiding the network to learn at different granularity levels. The features from these three different granularity levels are then fed into the Multi-scale Feature Aggregation Module (MFAM) for fusion. Finally, a Multi-Scale Regression Module (MSRM) is designed to handle the multi-scale fused features, resulting in crowd features rich in high-level semantics and low-level detail. Experimental results on five benchmark datasets demonstrate that TCHNet achieves highly competitive performance compared to some popular crowd counting methods.

Infrared point target detection based on multiscale homogeneous feature fusion

The added value of multiple data sources on tree species mapping has been widely analyzed. In particular, fusion of hyperspectral (HS) and LiDAR sensors for forest applications is a very hot topic. In this paper, we exploit the use of multi-scale features to fuse HS and LiDAR data for tree species mapping. Hyperspectral data is obtained from the APEX sensor with 286 spectral bands. LiDAR data has been acquired with a TopoSys sensor Harrier 56 at full waveform. We generate multi-scale features on both HS and LiDAR data, by considering the diameter and the height layer of different tree species. Experimental results on a forested area in Belgium demonstrate the effectiveness of using multi-scale features for fusion of HS image and LiDAR data both visually and quantitatively.