Aiming at the serious occlusion and slow tracking speed in pedestrian target recognition complex scenes, a method based on improved YOLO v5 combined with Deep SORT is proposed.By merging attention mechanism ECA-Net Neck part of network, using CIoU loss function non-maximum value suppression, connecting model Shuffle Net V2 as appearance feature extraction network to achieve lightweight fast purpose improving under occlusion.A large number experiments show that increases average precision by...

<abstract> <p>In response to the limited detection ability and low model generalization of YOLOv7 algorithm for small targets, this paper proposes a based on improved steel surface defect detection. First, Transformer-InceptionDWConvolution (TI) module is designed, which combines Transformer InceptionDWConvolution increase network's detect objects. Second, spatial pyramid pooling fast cross-stage partial channel (SPPFCSPC) structure introduced enhance network training...

Currently big data is becoming the worldwide focus of attention, and using machine learning techniques to obtain valuable information from massive complex structures has become a common concern yet an urgent problem.This paper analyzes summarizes present evaluation index under data, introduces some algorithms, then compares differences between traditional algorithms those explores its developing trend.

Patients with alcohol-induced psychiatric and behavioral disorders have higher drinking relapse rates after treatment when compared to those without these disorders.To investigate factors influencing among patients provide guidance for rehabilitative intervention being treated substance use disorders.Patients were randomly assigned into either the study group or control group. We used Chi-square test analyze their general demographics, history, hospitalizations. Factors analyzed by logistic...

The state-of-the-art trackers using deep learning technology have little special strategy to gain the bounding box well when target suffers drastic geometric deformation. In this paper, we take full use of convolutional neural network (CNN) features deepest layer represent semantic feature model, and affine transformation be as space information model. A tracking method based on geometrical region CNN is proposed. Firstly, applied predict possible locations a target, candidate boxes obtained...

For the problem that object tracking algorithm using principal components analysis(PCA) has low accuracy in a complex environment, based on partial least squares analysis, an is proposed by particle filtering with dual models. Firstly, model of region built which applied as observation model. Then, dynamic models Lie group and corresponding tangent vector space respectively, describing deformation process affine transformation. Finally, combining update strategy for feature space, realized...

Abstract Human pose estimation is an important task in computer vision, which can provide key point detection of human body and obtain bone information. At present, mainly utilized for large targets, there no solution small targets. This paper proposes a multi-channel spatial information feature based (MCSF-Pose) algorithm to address the issue medium targets inaccurate points scenarios involving occlusion multiple poses. The MCSF-Pose network bottom-up regression network. Firstly, UP-Focus...

To address the problem of dense crowd face detection in complex environments, this paper proposes a model named Deep and Compact Face Detection (DCFD), which adopts an improved lightweight EfficientNetV2 network to replace backbone RetinaFace. A large kernel attention mechanism is introduced task more accurately. The network, efficient channel (ECA) mechanism, added further improve algorithm performance. feature fusion module neural architecture search pyramid (NAS-FPN) that significantly...

In power system, information transmission pays more attention to security, stability and reliability. order meet the requirement of safety protection smart grid automation control, a scheme distribution based on FTU is proposed. Without changing current operation terminals, characteristics system are provided, such as wide range points, outdoor operation. And due industrial-level hardware design encryption chip, secure operating , SM1, SM2, SM3, SM4 national secret algorithm PKI (public key...

Compared with affine transformation, projection transformation represents the process of imaging objects more accurately. This paper proposes a novel object tracking method using particle filtering dual manifold models. One is covariance used for observation model, and other geometric deformation on SL (3) group, where rank matrix equals 1, adapted to utilize dynamic model. Our main contribution both geometry group manifolds in developing general filtering-based algorithm. Extensive...

For visual tracking, key factors that affect the performance of trackers are related to whether it can effectively extract appearance information and spatial a target. And most state-of-the-art either do not model separately or design special strategies deal with strong geometric deformation In this paper, we an separately, then combine them obtain complementary benefits. Firstly, because features from deeper layers convolutional neural network (CNN) better describe semantic target while...

Big Data is becoming the attentive focus in current world.With rapid integration and development of next generation information technology, such as Cloud computing, mobile Internet Things, data present an exponential growth.This paper illustrates concept big data, national international research application status, particularly analyzes advantages disadvantages key techniques processing, summarizes challenges that facing.Finally, it views prospects future based on above summaries.In recent...

Introduction The primary focus of this paper is to assess urban ecological environments by employing object detection on spatial-temporal data images within a city, in conjunction with other relevant information through mining. Methods Firstly, an improved YOLOv7 algorithm applied conduct detection, particularly counting vehicles and pedestrians the data. Subsequently, k-means superpixel segmentation utilized calculate vegetation coverage data, allowing for quantification area. This approach...

In the standard spectrum clustering algorithm, metric based on Euclidean space can not represent complicate distribution feature of some data set, which might lead to result inaccuracy.While geometric relationship between be describe more precise by manifold space.Considering Grassmann is a entropy Lie group, only has smooth curved surface but also fit for measuring distance data.All these make accurate.The improved analysis algorithm under Graasmann proposed this paper.The similarity...

In order to improve the quality and efficiency of image processing 3D reconstruction, a key frame screening method for video is proposed. Firstly, degradation performed by normalized sharpness evaluation function based on caused camera motion segmentation. Secondly, in avoid problem low accuracy triangulation, geometric robust information criterion GRIC was used as condition screen results shallow screening. The experimental show that three-dimensional reconstruction selected sequence source...

Visual object tracking methods based on wireless multimedia sensor network is one of the research hotspots while present linear method for processing feature vectors often lead to drift when with significant nonplanar pose variations through networks. In this article, we propose a novel nonlinear algorithm deformable objects. The proposed scheme has two filters. On hand, considering that Grassmann manifold entropy in Lie group manifold, which can describe and process data appearance more...

The state-of-the-art trackers using deep learning technology have no special strategy to capture the geometric deformation of target. Based on that affine manifold can better target shape change and higher level Convolutional Neural Network (CNN) describe semantic information objects, we propose a new tracking algorithm combining transformation with convolutional features track targets dramatic deformation. First, is applied predict possible locations target, then correlative filter designed...