Detection Of Dangerous Objects Research Articles

Detection of dangerous objects in images using neural networks

In recent years, along with the continuous improvement of the computer speed, the application of the computer for the examination and recognition of dangerous goods has become more and more widespread. In order to overcome the shortcomings of the high frequency of false detection in the process of classifying targets using existing feature-based classifiers, this paper proposes an algorithm for detecting dangerous objects using a convolutional neural network based on deep learning. For the image being checked, sliding windows of different scales are used to determine the presence of an object window. To detect objects, a convolutional neural network is trained with a large number of positive and negative samples. For better adaptation to object detection, the topology of the convolutional neural network has been improved. A window of suspected dangerous object is fed into an improved convolutional neural network to detect dangerous objects, and the false detection rate is reduced while maintaining the original detection rate.

Distribution of tasks in a clustered field of goals for homogeneous and heterogeneous UAV groups

This work is devoted to the distribution of tasks in groups of unmanned aerial vehicles (UAVs) under conditions of a significant excess of the number of tasks over the number of agents. The main tasks solved by UAVs: survey and reconnaissance of territories, detection of dangerous objects or places of emergencies, search for victims, etc. The efficiency of solving the problems listed above achieved by the simultaneous use of a group of UAVs, the ele-ments (agents) of which can carry out the tasks of inspecting and scanning various areas of space in parallel. The article proposes an iterative method for distributing tasks in a group of UAVs with a significant excess of the num-ber of tasks over the number of agents (5-20 times). The proposed method for heterogeneous groups of UAVs based on a two-stage procedure for distributing agents of different specializations among task clusters, taking into account the agent's value function. At the first stage, the base part of the agents is distributed, the remaining agents at the second stage distributed in order to average the distance traveled by each agent. Execution of tasks within a cluster implemented by simulating annealing. To evaluate the effectiveness of the method variants, a comparison made with the greedy task distribution algorithm and the collective goal distribution algorithm. The analogs under consideration are widespread, universal and have a high convergence of the solution. Experimental studies car-ried out by computer simulation, where 2000 experiments carried out with various changes in the number of group agents and generation of a task map. The results showed the high efficiency of the task distribution method in terms of reducing the distance traveled by the agents of the UAV group when performing tasks in comparison with analogues. The efficiency of the path traveled by agents is up to 28% depending on the number of agents and tasks in the cluster, which is a scientific increment of the result of the study.

A Compact W-Band Low-Noise Radiometry Sensor for a Single-Pixel Passive Millimeter-Wave Imager

Recently, studies on the remote detection of dangerous objects on the person have gained importance with increased security problems. Therefore, the use of passive millimeter waves in security systems is increasing because they are harmless to health and can penetrate clothes. In this study, a compact low-noise radiometric sensor (LNRS) that can be used to view hidden objects on the person was constructed. The LNRS can be arrayed thanks to its small size and ease of use, and can be used in imaging applications thanks to the 0.24 K resolution obtained. In addition, a passive millimeter imaging system (PMMWI) was developed to obtain images with the LNRS. The PMMWI system, which is realized in a quasi-optical structure, can be used in many experimental studies thanks to its compact structure.

Evaluation of the iterative method of task distribution in a swarm of unmanned aerial vehicles in a clustered field of targets

The purpose of this article is to evaluate the performance of the developed iterative method for distributing tasks in a swarm of UAVs in a clustered target field. This article presents an iterative method for distributing tasks among agents of a swarm of unmanned aerial vehicles (UAVs) in a clustered target field, when the number of tasks significantly exceeds the number of agents. The main tasks solved by UAVs are surveillance and reconnaissance, detection of dangerous objects or places of emergency, search for victims, etc. The efficiency of solving the problems listed above is achieved by the simultaneous use of a group of UAVs, the elements (agents) of which can carry out the tasks of inspecting and scanning various areas in parallel. At the same time, the number of tasks can significantly exceed the number of UAV swarm agents. To organize the work of a swarm of UAVs in this case, it is necessary to solve the problem of labor division, considering the conditions of the problem. The results showed the high efficiency of the proposed task distribution method according to the criterion of minimizing the travel distance of agents.

Selection of a given space object from the multitude observed by a specialized spacecraft

Introduction: For the detection of most dangerous artificial space objects (space junk) in near-earth space, it is planned to use specialized spacecraft equipped with optoelectronic devices. In relation to this a problem arose when selecting the most dangerous object from the observed multitude, based on the measurements of inhomogeneous selective features that characterize these objects. Purpose: To form a composite non-dimensional indicator that depends on the quantity and quality of measurement information about the observed space objects, and that determines the decision rule for selecting the most dangerous object to maximize the probability of making the right decision. Results: A method is proposed for selecting the most dangerous space object under the condition of the limited amount of measurement information about physically inhomogeneous selective features of space objects located in the area which is observed by a specialized spacecraft. It should be noted that the measurement data on individual selective features of space objects may be absent. The proposed decision rule for the selection of the most dangerous space object takes into account not only inaccuracies and errors, but also the number of measurements of the selective characteristics of each object. The efficiency of the method has been demonstrated on a relevant example. Practical relevance: The simplicity of the determination of composite indicators characterizing space objects that are located in the area observed by a specialized spacecraft, and of the decision rule for selecting the most dangerous object makes it possible to solve this problem on board a specialized spacecraft in real time.

Fast and accurate concealed dangerous object detection for millimeter-wave images

Millimeter-wave (MMW) imaging is a touch-free method for security inspection at railway stations and airports. However, automatic detection of dangerous objects in MMW images is challenging; the characteristics of images are remarkable with variable appearances of an object, different contrasts, and lower resolutions. We propose a fast and accurate concealed dangerous object detection method called MMW image Detection (MMWDet). MMWDet comprises two stages: single-image detection and multi-angle image detection. In the single-image detection stage, we design a more robust feature fusion relationship between adjacent feature maps in a feature pyramid network to achieve a more robust feature representation of an object. Meanwhile, we employ localization confidence rather than category confidence in the classification branch to address the negative sample interference in the training process. Then, multi-angle image detection is proposed to reduce the false and missed detection rate based on a single image by refining the detection results of multi-angle images, which maps and filters the prediction boxes frame by frame based on matching pairs of adjacent image feature points. We collected a MMW dataset that included 38k images. Extensive experiments on the MMW dataset demonstrate the effectiveness and improved performance of the proposed method compared with advanced general detection methods.

Low-Altitude Aerial Video Surveillance via One-Class SVM Anomaly Detection from Textural Features in UAV Images

In recent years, small-scale Unmanned Aerial Vehicles (UAVs) have been used in many video surveillance applications, such as vehicle tracking, border control, dangerous object detection, and many others. Anomaly detection can represent a prerequisite of many of these applications thanks to its ability to identify areas and/or objects of interest without knowing them a priori. In this paper, a One-Class Support Vector Machine (OC-SVM) anomaly detector based on customized Haralick textural features for aerial video surveillance at low-altitude is presented. The use of a One-Class SVM, which is notoriously a lightweight and fast classifier, enables the implementation of real-time systems even when these are embedded in low-computational small-scale UAVs. At the same time, the use of textural features allows a vision-based system to detect micro and macro structures of an analyzed surface, thus allowing the identification of small and large anomalies, respectively. The latter aspect plays a key role in aerial video surveillance at low-altitude, i.e., 6 to 15 m, where the detection of common items, e.g., cars, is as important as the detection of little and undefined objects, e.g., Improvised Explosive Devices (IEDs). Experiments obtained on the UAV Mosaicking and Change Detection (UMCD) dataset show the effectiveness of the proposed system in terms of accuracy, precision, recall, and F1-score, where the model achieves a 100% precision, i.e., never misses an anomaly, but at the expense of a reasonable trade-off in its recall, which still manages to reach up to a 71.23% score. Moreover, when compared to classical Haralick textural features, the model obtains significantly higher performances, i.e., ≈20% on all metrics, further demonstrating the approach effectiveness.

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.

Generative adversarial networks and faster-region convolutional neural networks based object detection in X-ray baggage security imagery

X-ray images have now been widely used in public transportation security. However, the ever-increasing number of dangerous object types, along with the complexity of the baggage's contents, makes it unsuitable for clearly displaying various baggages and hazards on X-ray images obtained directly from the inspection system. This article aims to improve the detection of dangerous objects by proposing a data augmentation method for enriching the X-ray prohibited item images using a Generative Adversarial Networks(GANs) based approach. Using the improved GANs model, a new X-ray images were generated with better quality and diversity, and the performance of the proposed method was evaluated using FID(Fréchet Inception Distance) score and Faster R-CNN (Region-based Convolutional Neural Network). Experimental results show that the new augmented X-ray security inspection image dataset can improve the detector performance. If other representative training datasets are utilized, we believe that our methodology could aid in the detection of other kinds of threat objects.

A YOLOv3 Based Smart City Application For Children’s Playgrounds

According to the reports of Public Health Institution, approximately 250,000 rabies-risk animal bites occur per year in Turkey. Most of these bites are caused by dogs and most of the victims are the children who play in playgrounds. With the development of deep learning-based computer vision technology, autonomous detection of dangerous objects (handguns, knives, dogs, etc.) in these children’s playgrounds has become a crucial security application. In this paper, a real-time dog detection model based on YOLOv3 deep learning algorithm is proposed as a new smart city security application and this model is applied to the selected children’s playground. Firstly, in view of the problem of insufficient stray dog image data in the original datasets, new images of stray dogs have been taken from an animal shelter and they have been added to the dataset. These new images have effectively enriched the diversity of training data and improved the training performance of the proposed model. The proposed model has been optimized by utilizing various hyperparameters and the results have been compared with each other. The model with the best evaluation scores is proposed and applied to detect dogs automatically by the fast emergency station located in the selected children’s playground. The real-time application has achieved 82.59% of AP with adjusted hyperparameters.

VIDEO SURVEILLANCE SYSTEMS FOR RAILWAY TRANSPORT

The article is devoted to analysis of ways to increase train traffic safety. One of ways to reduce accident rate on railways is timely detection of dangerous objects and notification of all traffic participants, primarily, the driver. Such a notification can be performed with the help of intelligent video surveillance system (IVS). The objective of the article is to study the possibility of using IVS to increase train traffic safety. Method (way to achieve the objective) consists of several stages: development of criteria of compliance of functional possibilities of IVS with tasks to increase train traffic safety, assessment, and calculation of permissible values of fitting criteria. The work suggests a scheme of organization of IVS based on fiber-optic data transmission system (FOTS) and data transmission network, highlights advantages and disadvantages of IVS, formulates requirements for IVS. Advantage of video surveillance is availability of video information about an object to a train driver, which allows to timely prevent an accident. Disadvantage of IVS is high probability of false detection, which can lead to false triggering of a system. To reduce the number of false triggering there are two ways: improvement of algorithms of video analytics in recognition device (RD) and increase of quality of video signals at the input of RD. The work is devoted to reduction of probability of false triggering due to improvement of communication quality. It is noted that an efficient method is the use of a new element base of electronics based on nanostructured materials. These materials allow to improve optoelectronic characteristics of main elements of IVS which are photosensors and photoreceivers.

A Multimodal Radar-Ultrasound Imaging Concept for Improved Display of Concealed Objects

Subsurface radar imaging provides an insight into hidden scenes: For example, in security screening, radar can be used for the detection of dangerous objects or contraband. In non-destructive testing, radar imaging also allows to find internal material defects such as cracks or contaminations. Object contours and internal structures appear as echoes in the radar receive signal. The strongest echo typically is the one from the test object's surface. This echo can mask smaller echoes close to the surface and therefore, small inner structures may become invisible in the reconstruction image. To overcome this problem, we propose a novel concept to eliminate the surface echo in radar imaging. In contrast to existing surface elimination techniques, our approach can be used not only when screening objects with planar surfaces, but also on such that exhibit an unknown and possibly non-planar profile. Our approach is to combine radar and air-coupled ultrasound. Since air-coupled ultrasound cannot penetrate into solids, the receive signal only contains the echo from the test object's surface. The core of our concept is to generate radar and ultrasound signals with the same characteristics, i.e. they are equal in amplitude and phase. Then, the ultrasound data can be subtracted from the radar data and the resulting signal will only contain the echoes of the inner structures. When processing this data, the reconstructed images are almost free of reflections from the surface.

Multi-task learning for dangerous object detection in autonomous driving

Recently, autonomous driving has been extensively studied and has shown considerable promise. Vision-based dangerous object detection is a crucial technology of autonomous driving. In previous work, dangerous object detection is generally formulated as a typical object detection problem and a distance-based danger assessment problem, separately. These two problems are usually dealt with using two independent models. In fact, vision-based object detection and distance prediction present prominent visual relationship. The objects with different distance to the camera have different attributes (pose, size and definition), which are very worthy to be exploited for dangerous object detection. However, these characteristics are usually ignored in previous work. In this paper, we propose a novel multi-task learning (MTL) method to jointly model object detection and distance prediction with a Cartesian product-based multi-task combination strategy. Furthermore, we mathematically prove that the proposed Cartesian product-based combination strategy is more optimal than the linear multi-task combination strategy that is usually used in MTL models, when the multi-task itself is not independent. Systematic experiments show that the proposed approach consistently achieves better object detection and distance prediction performances compared to both the single-task and multi-task dangerous object detection methods.

Locating the position of objects in non-line-of-sight based on time delay estimation**Project supported by the National Science and Technology Major Project of China (Grant No. AHJ2011Z001) and the Major Research Project of Yili Normal University (Grant No. 2016YSZD05).

Non-line-of-sight imaging detection is to detect hidden objects by indirect light and intermediary surface (diffuser). It has very important significance in indirect access to an object or dangerous object detection, such as medical treatment and rescue. An approach to locating the positions of hidden objects is proposed based on time delay estimation. The time delays between the received signals and the source signal can be obtained by correlation analysis, and then the positions of hidden objects will be located. Compared with earlier systems and methods, the proposed approach has some modifications and provides significant improvements, such as quick data acquisition, simple system structure and low cost, and can locate the positions of hidden objects as well: this technology lays a good foundation for developing a practical system that can be used in real applications.

Application of Magnetovision Scanning System for Detection of Dangerous Objects

This paper presents application of magnetovision scanning system for detection of dangerous objects in weak magnetic fields. Measurement system was developed to study the magnetic field vector distributions. The measurements of the Earth’s field disturbances caused by dangerous ferromagnetic objects were carried out. The ability for passive detection of hidden dangerous objects and determine their location was demonstrated.

Magnetovision Scanning System for Detection of Dangerous Objects

This paper presents an application of magnetovision based measurements to develop a method for passive detection of dangerous ferromagnetic objects. Scanning system was designed and built to study the magnetic eld vector distributions. The measurements of the Earth's eld disturbances caused by ferromagnetic objects were carried out. The ability for passive detection of selected dangerous objects was demonstrated. Further data processing allowed for determining the (x,y) coordinates of the object relative to the plane of measurement, and even the possibility of calculating the distance from the object. The results obtained indicate that it is possible to detect and determine the location of dangerous ferromagnetic objects. This opens the new way to use magnetovision in public security systems, in particular for the detection of dangerous objects by police and sapper robots.

Ferromagnetic Objects Magnetovision Detection System.

This paper presents the application of a weak magnetic fields magnetovision scanning system for detection of dangerous ferromagnetic objects. A measurement system was developed and built to study the magnetic field vector distributions. The measurements of the Earth’s field distortions caused by various ferromagnetic objects were carried out. The ability for passive detection of hidden or buried dangerous objects and the determination of their location was demonstrated.

The method of multifrequency near-field acoustical tomography of bulk inhomogeneities of the sea bottom

In this work, we propose a method of near-field acoustical multi-frequency coherent tomography of spatially localized inhomogeneities of the shallow-sea bottom. In the framework of the developed method, two-dimensional spatial scanning of the radiating-receiving system along the bottom over the region of location of the inhomogeneities is performed at multiple frequencies. Using the Born approximation, the initial three-dimensional integral equation for the scattered field is reduced to one-dimensional Fredholm equation of the first kind with respect to the depth profile of the transverse spatial spectrum of the inhomogeneities. In the solution of this integral equation for each pair of spectral components, we use the method of generalized residual and obtain the sought three-dimensinal distribution via the inverse Fourier transform of the reconstructed spectrum. Results of numerical simulation of the tomography scheme and visualized inhomogeneities of shallow-sea bottom are presented. Acoustical methods of remote sensing of the sea are an efficient tool in solving various practical problems related to the development of sea-shelf regions [1–3]. In recent years, remote acoustic sensing of the sea bottom has become especially topical. This is related to solving many practical problems of engineering seismics, detection of buried objects in sedimentary layers, and ecological monitoring in the regions of ports, underwater pipelines, and oil-producing platforms. For solving such problems, pulse-type methods using non-coherent seismo-acoustic sources of explosive type became widely spread. Along with this, promising methods of coherent sounding of the sea bottom using towed hydroacoustic radiators are also studied. Utilization of such sources makes it possible to coherently accumulate signals, which significantly helps to increase sensitivity and resolution in the sounding problems [4–8]. However, usually the spatial processing is performed only for synthesized apertures which are formed by translational one-directional motion of the radiating and receiving systems. Creation and testing of coherent hydroacoustic radiators for sea-bottom sounding makes it possible to apply even more efficient methods using the data of two- and three-dimensional scanning [1, 2]. We note that many of such diagnostic problems based on data of multi-dimensional observations (see, for example, [8– 10]) belong to the class of ill-posed inverse problems, the solution of which can be based on application of regularization methods [9, 10] and use of ap rioriinformation [11]. Such problems are topical for underwater engineering prospecting and detection of dangerous objects buried in sedimentary layers. The tomographic method proposed in the present work, i.e., the method of reconstruction of threedimensional structure of inhomogeneities located in the sea bottom based on solving the corresponding backscattering problem, is a natural development of the above-mentioned research. Studies and applications of backscattering problems have a long history not only in acoustics, but also in electrodynamics and quantum mechanics. The simplest problem of such kind, for an infinite one-dimentionally inhomogeneous