Surveillance Tasks Research Articles

Real-time pedestrian detection method based on improved YOLOv3

Pedestrian detection in image or video data is a very important and challenging task in security surveillance. The difficulty of this task is to locate and detect pedestrians of different scales in complex scenes accurately. To solve these problems, a deep neural network (RT-YOLOv3) is proposed to realize real-time pedestrian detection at different scales in security monitoring. RT-YOLOv3 improves the traditional YOLOv3 algorithm. Firstly, the deep residual network is added to extract vehicle features. Then six convolutional neural networks with different scales are designed and fused with the corresponding scale feature maps in the residual network to form the final feature pyramid to perform pedestrian detection tasks. This method can better characterize pedestrians. In order to further improve the accuracy and generalization ability of the model, a hybrid pedestrian data set training method is used to extract pedestrian data from VOC data set and train with INRIA pedestrian data set. Experiments show that the proposed RT-YOLOv3 method achieves 93.57% accuracy of mAP (mean average precision) and 46.52f/s (number of frames per second). In terms of accuracy, RT-YOLOv3 performs better than Fast R-CNN, Faster R-CNN, YOLO, SSD, YOLOv2 and YOLOv3. This method reduces the missed detection rate and false detection rate, improves the positioning accuracy, and meets the requirements of real-time detection of pedestrian objects.

A Smart, Efficient, and Reliable Parking Surveillance System With Edge Artificial Intelligence on IoT Devices

Cloud computing has been a main-stream computing service for years. Recently, with the rapid development in urbanization, massive video surveillance data are produced at an unprecedented speed. A traditional solution to deal with the big data would require a large amount of computing and storage resources. With the advances in Internet of things (IoT), artificial intelligence, and communication technologies, edge computing offers a new solution to the problem by processing the data partially or wholly on the edge of a surveillance system. In this study, we investigate the feasibility of using edge computing for smart parking surveillance tasks, which is a key component of Smart City. The system processing pipeline is carefully designed with the consideration of flexibility, online surveillance, data transmission, detection accuracy, and system reliability. It enables artificial intelligence at the edge by implementing an enhanced single shot multibox detector (SSD). A few more algorithms are developed on both the edge and the server targeting optimal system efficiency and accuracy. Thorough field tests were conducted in the Angle Lake parking garage for three months. The experimental results are promising that the final detection method achieves over 95% accuracy in real-world scenarios with high efficiency and reliability. The proposed smart parking surveillance system can be a solid foundation for future applications of intelligent transportation systems.

Implementation of an Adaptive Artificial Vision System Applied to Video Surveillance Tasks on Raspberry Pi

Implementation of an Adaptive Artificial Vision System Applied to Video Surveillance Tasks on Raspberry Pi

ЗАСТОСУВАННЯ НАПІВАКТИВНИХ РАДІОЛОКАЦІЙНИХ СИСТЕМ (У ПОЛІ ПІДСВІЧУВАННЯ СТОРОННІХ ДЖЕРЕЛ) ДЛЯ ВИКОНАННЯ ЗАВДАНЬ РОЗВІДКИ ПОВІТРЯНОГО ПРОСТОРУ

У статті розглянута можливість використання сигналів покриття мережі інтернет, стільникових телефонних мереж, сигналів радіомовлення та телебачення для забезпечення виявлення повітряних цілей. Розглянуто переваги, недоліки, склад та принцип роботи напівактивної радіолокаційної ситеми. Досліджені питання технічної реалізації антен та приймачів напівактивної радіолокаційної системи. Запропоновано алгоритм обробки сигналів у напівактивній радіолокаційній системі. Доповнення системи радіолокаційної розвідки повітряного простору засобами напівактивної радіолокації забезпечить її скритність, а відповідно й живучість.

Engineering and Deploying a Cheap Recognition Security System on a Raspberry Pi Platform for a rural Settlement

Security is one of the most fundamental challenges of mankind, providing affordable devices for apprehending criminals. Using smart technology is on the rise and the ability to have full surveillance records of both authorised and unauthorized entrance to designated facility or important resource in a timely manner is highly desirable in modern society of today. This paper proposes the use of Histogram of Oriented Gradients (HOG) to train a model capable of recognising authorised personnel on a raspberry pi device for the purpose of security and ease of access to vital infrastructure. HOG was the preferred choice because it is not computationally intensive as compared to Convolutional Neural Networks (CNN) and most other relatively comparable computational algorithms. The HOG network detect faces and sends a report to Firebase Database and an image is also sent to Google Cloud Storage (GCS) a package on the Google Cloud Platform (GCP). Both data from Firebase and GCS are sent to a companion android application where the user can view who entered specific locations, at specific time with accompanying pictorial evidence. The recognition system was deployed on a raspberry pi device that’s feeds in visual data via an inexpensive camera. Collectively, the proposed system is a relatively cheap smart technology security system with inherent ability to accomplish real-time surveillance tasks using widely penetrated android phone technology while maintaining low computational overheads.

Learning a Layout Transfer Network for Context Aware Object Detection

We present a context aware object detection method based on a retrieve-and-transform scene layout model. Given an input image, our approach first retrieves a coarse scene layout from a codebook of typical layout templates. In order to handle large layout variations, we use a variant of the spatial transformer network to transform and refine the retrieved layout, resulting in a set of interpretable and semantically meaningful feature maps of object locations and scales. The above steps are implemented as a Layout Transfer Network which we integrate into Faster RCNN to allow for joint reasoning of object detection and scene layout estimation. Extensive experiments on three public datasets verified that our approach provides consistent performance improvements to the state-of-the-art object detection baselines on a variety of challenging tasks in the traffic surveillance and the autonomous driving domains.

Deep Learning-Based Stair Segmentation and Behavioral Cloning for Autonomous Stair Climbing

Mobile robots are widely used in the surveillance industry, for military and industrial applications. In order to carry out surveillance tasks like urban search and rescue operation, the ability to traverse stairs is of immense significance. This paper presents a deep learning-based approach for semantic segmentation of stairs, behavioral cloning for stair alignment, and a novel mechanical design for an autonomous stair climbing robot. The main objective is to solve the problem of locomotion over staircases with the proposed implementation. Alignment of a robot with stairs in an image is a traditional problem, and the most recent approaches are centered around hand-crafted texture-based Gabor filters and stair detection techniques. However, we could arrive at a more scalable and robust pipeline for alignment schemes. The proposed deep learning technique eliminates the need for manual tuning of parameters of the edge detector, the Hough accumulator and PID constants. The empirical results and architecture of stair alignment pipeline are demonstrated in this paper.

Prospective memory in the red zone: Cognitive control and capacity sharing in a complex, multi-stimulus task.

Remembering to perform a planned action upon encountering a future event requires event-based Prospective Memory (PM). PM is required in many human factors settings in which operators must process a great deal of complex, uncertain information from an interface. We study event-based PM in such an environment. Our task, which previous research has found is very demanding (Palada, Neal, Tay, & Heathcote, 2018), requires monitoring ships as they cross the ocean on a display. We applied the Prospective Memory Decision Control Model (Strickland, Loft, Remington, & Heathcote, 2018) to understand the cognitive mechanisms that underlie PM performance in such a demanding environment. We found evidence of capacity sharing between monitoring for PM items and performing the ongoing surveillance task, whereas studies of PM in simpler paradigms have not (e.g., Strickland et al., 2018). We also found that participants applied proactive and reactive control (Braver, 2012) to adapt to the demanding task environment. Our findings illustrate the value of human factors simulations to study capacity sharing between competing task processes. They also illustrate the value of cognitive models to illuminate the processes underlying adaptive behavior in complex environments. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Cross-Modality Person Re-Identification Based on Dual-Path Multi-Branch Network

Person re-identification is an important surveillance task of searching and identifying pedestrian across different images or video frames. Despite a significant progress has been made in person re-identification based on RGB image sensors, few work focus on the person re-identification between RGB and infrared images, which is a challenging cross-modality problem and has been widely encountered in a dark environment or at night. In addition to the challenges for the same identity associated with variations in camera viewpoints and person poses, there is a non-negligible shift across different sensor modalities since the visual characteristics from RGB and infrared images are heterogeneous. In this paper, we propose a novel end-to-end dual-path multi-branch network for RGB-infrared cross-modality person re-identification, which introduces the multi-branch deep network architecture. The experimental results obtained with SYSU-MM01 datasets indicate that the proposed method can successfully transfer descriptive visual characteristic between RGB and infrared sensor modality. It can significantly outperform state-of-the-art conventional methods and convolutional neural network methods.

Incorporation of Contingency Tasks in Task Allocation for Multirobot Teams

Complex logistics support missions require the execution of spatially separated information gathering and situational awareness tasks. Mobile robot teams can play an important role in the automated execution of these tasks to reduce mission completion time. Planning strategies for such missions must take into account the formation of effective coalitions among available robots and assignment of tasks to robots with the goal of minimizing the expected mission completion time. The occurrence of unexpected situations that adversely interfere with the execution of the mission may require the execution of contingency tasks so that the originally planned tasks may proceed with minimal disruption. Initially reported potential contingency tasks may not always affect mission tasks due to the uncertainty in the mission environment. When potential contingency tasks are reported, the planner updates its existing plan to minimize the expected mission completion time based on the probability of these contingency tasks impacting the mission, their impact on the mission, and other task characteristics. We describe various heuristic-based strategies to compute task allocations for robots for mission execution. We perform simulation experiments to compare them and analyze the computational performance of the best performing strategy. We show that the proactive approach to contingency task management outperforms both the conservative and reactive approaches. Note to Practitioners —The work reported in this article will be useful for deploying multirobot teams to support complex logistics missions spread over a large area where the robots must be prepared to handle contingencies that can adversely impact the mission. The proposed proactive approach can be used to handle contingencies in information gathering, surveillance, guarding, and situational awareness tasks to support safe and secure transportation of important assets through crowded areas. We use port operation as an illustrative example where unmanned surface and aerial vehicles can be useful in ensuring the safety and security of the ports. This is a computationally challenging problem. This article proposes heuristic algorithms to solve the task allocation problem among many different agents efficiently. The approach presented in this article integrates the available information regarding mission and contingencies, along with the resource constraints to plan the mission execution.

The Repair Strategy for Event Coverage Holes Based on Mobile Robots in Wireless Sensor and Robot Networks.

In the application of the wireless sensor and robot networks (WSRNs), there is an urgent need to accommodate flexible surveillance tasks in intricate surveillance scenarios. On the condition of flexible surveillance missions and demands, event coverage holes occur in the networks. The conventional network repair methods based on the geometric graph theory such as Voronoi diagram method are unable to meet the conditions of flexible surveillance tasks and severe multi-restraint scenarios. Mobile robots show obvious advantages in terms of adaptation capacity and mobility in hazardous and severe scenarios. First, we propose an event coverage hole healing model for multi-constrained scenarios. Then, we propose a joint event coverage hole repair algorithm (JECHR) on the basis of global repair and local repair to apply mobile robots to heal event coverage holes in WSRNs. Different from conventional healing methods, the proposed algorithm can heal event coverage holes efficaciously which are resulted from changing surveillance demands and scenarios. The JECHR algorithm can provide an optimal repair method, which is able to adapt different kinds of severe multi-constrained circumstances. Finally, a large number of repair simulation experiments verify the performance of the JECHR algorithm which can be adapted to a variety of intricate surveillance tasks and application scenarios.

ZIKA: improved surveillance and forecast of Zika virus in Brazil

Abstract Background The devastating consequences in neonates infected with Zika virus make controlling the spread of the virus and its vectors (Aedes mosquitoes) an urgent challenge. Mobile technology has been successfully used to support routine surveillance of mosquito populations conducted by the community health workers (health agents). Mobile real-time surveillance data also provides a ground truth for calibrating early-warning and prediction systems. Methods The ZIKA app supports the health agents in routine surveillance of properties with risk of mosquito breeding, recommends the most optimal route to visit the assigned properties, and includes gamification features, such as badges, to reward healthcare agents for successfully completing the surveillance tasks. Secondly, the project developed a model of the vector population spread using neural networks and artificial intelligence methods combing data from healthcare agents and mosquito ovitraps from the period 2009-2014, with climatic and weather data, calibrated by the real time data from the ZIKA app. Results The results from testing sessions demonstrated the app could replace the very time consuming paper-based systems and provide an improved surveillance data available at real time, for seamless reporting to the regional authorities. Using the 2009-2014 dataset, the neural network model successfully predicts mosquito occurrence in the high risk regions of the city of Recife. Conclusions The ZIKA mobile app provides a much improved surveillance of the mosquito populations and real-time data for the public health authorities responsible for rapid response. The AI early warning model successfully predicts mosquito populations using healthcare agents data, weather and climatic data. Key messages ZIKA project developed a mobile app to improve surveillance of mosquito population in tropical regions of Brazil. ZIKA early warning system, using AI neural networks, has the potential to predict the risk of mosquito populations in high risk urban locations.

Hull Shape Actuation for Speed Regulation in an Underwater Vehicle

Abstract Autonomous underwater vehicles (AUVs) have shown great promise in fulfilling surveillance, scavenging, and monitoring tasks. Traditional gliders and streamlined AUVs are designed for long-term operational efficiency in expansive environments but are limited in cluttered spaces due to their shape and control authority; agile AUVs can penetrate cluttered or sensitive environments but are limited in operational endurance at large spatial scales. This paper presents the dynamic modeling and control design of an underwater hull capable of actuating its shape morphology. The prototype hull incorporates flexible, buckled fiberglass ribs to ensure a rigid shape that is actuated by modulating the length of the body’s semi-major axis. We represent the vehicle shape using a single control input actuating the vehicle’s length-to-diameter ratio. Hydrodynamic modeling of the flexible hull suggests that dynamic shape actuation can modulate the mass and drag properties of the hull to improve control of the vehicle’s forward speed. Using tools from nonlinear control theory, this paper presents the derivation of a shape-actuating autonomous control algorithm regulating the vehicle speed to a time-varying reference speed, subject to the actuator limits. The theoretical control results are illustrated using numerical simulations of the vehicle model.

Behavior-Based Control for an Aerial Robotic Swarm in Surveillance Missions.

Aerial robotic swarms have shown benefits for performing search and surveillance missions in open spaces in the past. Among other properties, these systems are robust, scalable and adaptable to different scenarios. In this work, we propose a behavior-based algorithm to carry out a surveillance task in a rectangular area with a flexible number of quadcopters, flying at different speeds. Once the efficiency of the algorithm is quantitatively analyzed, the robustness of the system is demonstrated with 3 different tests: loss of broadcast messages, positioning errors, and failure of half of the agents during the mission. Experiments are carried out in an indoor arena with micro quadcopters to support simulation results. Finally, a case study is proposed to show a realistic implementation in the test bed.

Automatic Change Detection System over Unmanned Aerial Vehicle Video Sequences Based on Convolutional Neural Networks.

In recent years, the use of unmanned aerial vehicles (UAVs) for surveillance tasks has increased considerably. This technology provides a versatile and innovative approach to the field. However, the automation of tasks such as object recognition or change detection usually requires image processing techniques. In this paper we present a system for change detection in video sequences acquired by moving cameras. It is based on the combination of image alignment techniques with a deep learning model based on convolutional neural networks (CNNs). This approach covers two important topics. Firstly, the capability of our system to be adaptable to variations in the UAV flight. In particular, the difference of height between flights, and a slight modification of the camera’s position or movement of the UAV because of natural conditions such as the effect of wind. These modifications can be produced by multiple factors, such as weather conditions, security requirements or human errors. Secondly, the precision of our model to detect changes in diverse environments, which has been compared with state-of-the-art methods in change detection. This has been measured using the Change Detection 2014 dataset, which provides a selection of labelled images from different scenarios for training change detection algorithms. We have used images from dynamic background, intermittent object motion and bad weather sections. These sections have been selected to test our algorithm’s robustness to changes in the background, as in real flight conditions. Our system provides a precise solution for these scenarios, as the mean F-measure score from the image analysis surpasses 97%, and a significant precision in the intermittent object motion category, where the score is above 99%.

Optimization of Shadow Detection and Removal using Multilevel Thresholds and Improved Artificial Bee Colony Algorithm

Shadow Detection and removal from images is a challenging task in visual surveillance and computer vision applications. The appearance of shadows creates severe problems. There are various methods already exists but scope in this area is wide and open. In this paper, Optimization of Shadow Detection and Removal using Improved Artificial Bee Colony Algorithm (IABC) is proposed. The proposed method uses edge map, multilevel thresholds, masking, boundaries evaluation and, IABC algorithm. First data pre-processing is applied to find the correlation between the pixels then three level low, medium and high value of thresholds and the corresponding value of masking and boundaries are calculated to accurately differentiate pixels as foreground. The edge response, curvature, gradient are applied to find the true location of boundaries. Finally, IABC has been applied for detecting the shadow and median filter is used to remove the shadow. The results show improvement as compared to other existing methods.

Unmanned aerial vehicle-assisted energy-efficient data collection scheme for sustainable wireless sensor networks

Recently, wireless sensor networks (WSNs), consisted of battery-powered sensor nodes, are widely adopted by various civilian/military applications for implementing real-time monitoring or long-term surveillance tasks. However, due to limited battery capacity, sensor lifetime is limited, which in turn affects the working time of the whole system. Therefore, solving the problem of improving energy efficiency is the premise for constructing a sustainable/energy-efficient WSN. To address this problem, by taking advantage of the emerging unmanned aerial vehicle (UAV) techniques, we propose a novel UAV-assisted data collection scheme for improving the energy efficiency of the deployed WSN. To meet the performance requirements of systems with different scales, our algorithm has two working modes: single- and multiple-UAV scenarios. For the small-scale system, a single UAV considered data-collector is adopted to access and collect data from the deployed nodes. For the large-scale system, based on a newly designed clustering method, multiple data collectors are utilized to gathering sensed data from the deployed nodes for conserving their energy and reducing the data forwarding latency. To evaluate our proposed work, intensive simulation experiments are conducted. The results demonstrate that the proposed algorithm achieved better performance than the control group regarding system-wide energy efficiency and average end-to-end delay.

Detection of influenza virus by agglutination using nanoparticles conjugated with a sialic acid-mimic peptide

Influenza virus (IFV) detection in the early phase of disease is critical for effective anti-influenza therapy using neuraminidase inhibitors. Sialyloligosaccharide receptors on the surface of respiratory cells are recognized by IFV hemagglutinin (HA) in the infection. Here, we show that agglutination of IFV is detected using poly(glycidyl methacrylate) (PGMA)-coated polystyrene nanoparticles conjugated with a sialic acid-mimic peptide. The azido peptide was immobilized onto the surface of the PGMA-coated nanoparticles by click chemistry. The distribution of particle size, determined by dynamic light scattering, indicated that the peptide-conjugated nanoparticles were agglutinated in the presence of HA and IFV. Nanoparticles conjugated with the receptor-mimic peptide may be a useful alternative to red blood cells in the global surveillance and clinical diagnosis of influenza. This study examines the use of poly(glycidyl methacrylate) (PGMA)-coated polystyrene functionalized with peptides that mimic receptors found on red blood cells for detection of the influenza virus. Although the current method, the hemagglutination inhibition test, utilized by the World Health Organization, is simple, it has some limitations. We found that our peptide dimers attached to beads were efficiently agglutinated, leading us to detect the presence of the influenza virus antigens. We believe that our study makes a significant contribution to the fight against influenza because we provide an alternative method for the important task of influenza surveillance and diagnosis.

Anxious individuals predict the onset of aggression earlier in a CCTV surveillance task.

Attentional biases in anxious individuals can facilitate the detection of threatening stimuli. A particular field of research that may benefit from enhanced threat detection is in closed-circuit television (CCTV) surveillance, in which operators search through multiple camera feeds to attempt to identify threatening situations before they occur. The present study examined whether the enhanced threat detection of anxious individuals extends to the ability to detect threat in a multiple-scene CCTV task. Anxiety was measured in a nonclinical sample using the State-Trait Inventory for Cognitive and Somatic Anxiety. Participants were asked to try to find aggressive incidents that were simulated using the game Grand Theft Auto V in displays showing 1, 4, or 9 simultaneous videos. The results revealed that higher levels of trait cognitive, state cognitive, and trait somatic anxiety were related to earlier responses, with no change in confidence or accuracy. Increasing the number of screens to be monitored was associated with detecting the events later and a reduced confidence in responses. These results suggest that, in nonclinical populations, a moderate degree of anxiety may be beneficial in predicting acts of aggression during CCTV monitoring tasks. Trait (i.e., stable) levels of anxiety may inform recruitment of operators for surveillance tasks. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Comprehensive Analysis of Deep Learning-Based Vehicle Detection in Aerial Images

Vehicle detection in aerial images is a crucial image processing step for many applications such as screening of large areas as used for surveillance, reconnaissance, or rescue tasks. In recent years, several deep learning-based frameworks have been proposed for object detection. However, these detectors were developed for data sets that considerably differ from aerial images. In this paper, we systematically investigate the potential of fast R-CNN and faster R-CNN for aerial images, which achieve top performing results on common detection benchmark data sets. Therefore, the applicability of eight state-of-the-art object proposal methods used to generate a set of candidate regions and of both detectors is examined. Relevant adaptations to account for the characteristics of the aerial images are provided. To overcome the shortcomings of the original approach in the case of handling small instances, we further propose our own networks that clearly outperform state-of-the-art methods for vehicle detection in aerial images. Furthermore, we analyze the impact of the different adaptations with respect to various ground sampling distances to provide a guideline for detecting small objects in aerial images. All experiments are performed on two publicly available data sets to account for differing characteristics such as varying object sizes, number of objects per image, and varying backgrounds.