Vision-based Approach Research Articles

Development and evaluation of a vision-based transfer learning approach for indoor fire and smoke detection

Fire poses a significant risk across industrial and domestic settings, especially to firefighters who must tackle the blaze. Current technology for detection in indoor environments are smoke detectors and flame detectors. However, these detectors have several limitations during the ignition phase of a fire and propagation. These systems cannot detect an exact position of the fire nor how the fire is spreading or its size, all of which is necessary information for fire services when dealing with these incidents. A potential solution is to use artificial intelligence techniques such as computer vision, which has shown the potential to detect and recognise objects and activities in indoor spaces. This study aims to develop a vision-based fire and smoke detection system. A deep learning technique that incorporates convolutional neural networks (CNN) was utilised to develop the real-time detection approach that can potentially provide necessary information for fire services, including identifying the position and size of the fire and how the fire spreads. A transfer learning approach using a pre-trained model was used to train the detector. Based on the detection and recognition tests using indoor fire and smoke videos, results indicated that the fire detection achieved up to 92.37% correct detections while the smoke detection did not perform as well. Hence, further improvement and evaluation of the detection approach will be conducted in future work, focusing on the impact of different parameters such as the detection model, building type, indoor space size and positioning of the detection camera. The present study provides an insight into the capabilities and potential applications of the concept.

A computer vision-based method for bridge model updating using displacement influence lines

This paper presents a new computer vision-based method that simultaneously provides the moving vehicle’s tire loads, the location of the loads on a bridge, and the bridge’s response displacements, based on which the bridge’s influence lines can be constructed. The method employs computer vision techniques to measure the displacement influence lines of the bridge at different target positions, which is then later used to perform model updating of the finite element models of the monitored structural system.The method is enabled by a novel computer vision-based vehicle weigh-in-motion method which the co-authors recently introduced. A correlation discriminating filter tracker is used to estimate the displacements at target points and the location of single or multiple moving loads, while a low-cost, non-contact weigh-in-motion technique evaluates the magnitude of the moving vehicle loads.The method described in this paper is tested and validated using a laboratory bridge model. The system was loaded with a vehicle with pressurized tires and equipped with a monitoring system consisting of laser displacement sensors, accelerometers, and cameras. Both artificial and natural targets were considered in the experimental tests to track the displacements with the cameras and yielded robust results consistent with the laser displacement measurements.The extracted normalized displacement influence lines were then successfully used to perform model updating of the structure. The laser displacement sensors were used to validate the accuracy of the proposed computer vision-based approach in deriving the displacement measurements, while the accelerometers were used to derive the system’s modal properties employed to validate the updated finite element model. As a result, the updated finite element model correctly predicted the bridge’s displacements measured during the tests. Furthermore, the modal parameters estimated by the updated finite element model agreed well with those extracted from the experimental modal analysis carried out on the bridge model.The method described in this paper offers a low-cost non-contact monitoring tool that can be efficiently used without disrupting traffic for bridges in model updating analysis or long-term structural health monitoring.

Recent advances in vision-based indoor navigation: A systematic literature review

Indoor navigation has remained an active research area for the last decade. Unlike outdoor environments, indoor environments have additional challenges, such as weak signals, low light, and complex scenarios. Different technologies are used for indoor navigation, including WiFi, Bluetooth, inertial sensors, and computer cameras. Vision-based methods have great potentials for indoor navigation as they fulfill most of the general requirements such as minimal cost, ease of use, ease of implementation, and realism. Therefore, researchers have successively proposed different novel vision-based approaches for indoor navigation. Unfortunately, there is no standard review article (except a few general reviews) that covers the current trends and draws a pipeline for future research. In this paper, we reviewed the current state-of-the-art vision-based indoor navigation methods. We followed the systematic literature review (SLR) methodology for article searching, selection, and quality assessments. In total, we selected 68 articles after final selection using SLR. We classified these articles into different categories. Each article is briefly studied for information extraction, including key idea, category of the article, evaluation criterion, and its strengths and weaknesses. We also highlighted several interesting future directions. This study will help new researchers to grasp the research challenge as well as present the results of their research in the field. It will also help the community to find a suitable indoor navigation system according to users’ requirements.

A vision-based hand hygiene monitoring approach using self-attention convolutional neural network

Automatic hand hygiene monitoring is an effective and necessary practice in healthcare. The main objective is to automate the quality control of the hand hygiene process by recognizing handwashing actions. This paper proposes a new computer vision-based hand hygiene monitoring approach, with the aim to recognize handwashing actions from the input videos. The proposed approach consists of the following three modules. First, in the feature extraction module, a backbone convolutional neural network (CNN) model (i.e., a ResNet model) is used to extract features from each frame of the input video. Second, in the feature aggregation module, the self-attention mechanism is introduced to form six masked self-attention blocks that are stacked to aggregate features from multiple frames. Third, in the feature classification module, the self-attentive aggregated feature representation is used for handwashing action recognition. The proposed approach outperforms a few state-of-the-art handwashing action recognition approaches. The proposed approach is able to perform effective and accurate handwashing action recognition for automatic hand hygiene monitoring, as verified by three benchmark datasets.

Graph-based linguistic and visual information integration for on-site occupational hazards identification

Construction sites are hazardous with various potential hazards that can occur at any time. The combination of different factors always causes the construction fatalities, and the majority of these fatalities could be prevented if workers followed on-site regulatory rules. However, compliance of regulatory rules is not strictly enforced among workers due to all kinds of reasons. Although previously proposed vision-based approaches are available for occupational hazards identification, the practicality is limited by the lack of automated understanding and adaptability to regulatory rules changes. In response to these gaps, this paper proposes a novel graph-based framework that integrates linguistic and visual information to process regulatory rule sentences and images for on-site occupational hazards identification. Particularly, a regulatory rules processing approach is presented to automatically extract and represent the key linguistic information of regulatory rules and a vision-based image scene information understanding approach is introduced to process on-site images by the combination of deep learning-based object detection and individual detection using geometric relationships analysis. Additionally, an automated reasoning approach is proposed to provide the integration of the processed linguistic and visual information and perform hazards identification. The hazards of two scenes, i.e., “working on height” and “operating a grinder”, were successfully identified with significantly higher performance compared to the baseline model.

Structural Health Monitoring of Large Structures via mmWave Sensing

Abstract Vibration and deformation measurement are of great significance for structural health monitoring (SHM). While the existing accelerometer, laser and vision-based approaches have limitations in multi-point vibration measurement, efficiency and environmental adaptability, especially for large structures. In this paper, we present a method for structural health monitoring with multi-point vibration and deformation measurement via millimetre- wave sensing. First, we describe the mathematical model and the principle of multi-target synchronous vibration and deformation monitoring. Second, with a built prototype of mmWave sensing system, we evaluate our method with both laboratory and field tests. The experimental results of different amplitudes, frequencies and velocities show that our approach can accurately monitor the vibration and deformation of multi-point with μm-level accuracy. The field test of Shanghai Songpu Bridge is provided to further validate the effectiveness of the method, offering a novel approach for contactless vibration and deformation monitoring of large structures in SHM.

User-oriented Natural Human-Robot Control with Thin-Plate Splines and LRCN

We propose a real-time vision-based teleoperation approach for robotic arms that employs a single depth-based camera, exempting the user from the need for any wearable devices. By employing a natural user interface, this novel approach leverages the conventional fine-tuning control, turning it into a direct body pose capture process. The proposed approach is comprised of two main parts. The first is a nonlinear customizable pose mapping based on Thin-Plate Splines (TPS), to directly transfer human body motion to robotic arm motion in a nonlinear fashion, thus allowing matching dissimilar bodies with different workspace shapes and kinematic constraints. The second is a Deep Neural Network hand-state classifier based on Long-term Recurrent Convolutional Networks (LRCN) that exploits the temporal coherence of the acquired depth data. We validate, evaluate and compare our approach through both classical cross-validation experiments of the proposed hand state classifier; and user studies over a set of practical experiments involving variants of pick-and-place and manufacturing tasks. Results revealed that LRCN networks outperform single image Convolutional Neural Networks; and that users' learning curves were steep, thus allowing the successful completion of the proposed tasks. When compared to a previous approach, the TPS approach revealed no increase in task complexity and similar times of completion, while providing more precise operation in regions closer to workspace boundaries.

RGB camera-based fallen person detection system embedded on a mobile platform

Most injuries in the elderly are due to falls. The response time, to attend to the critically injured in such fall cases, is crucial to their survival. This paper presents a low-cost, autonomous assistive patrol robot which additionally includes a fallen person detection module with facial recognition that allows identification of patients. Patrol robots could be beneficial for care centers, where there is a considerable number of patients that require care. In these conditions, falls can be generally detected by the robotic platform during the post-fall phase. This allows the system to work with no frame rate constraints, allowing other tasks to be run simultaneously. Based on the YOLO network, we propose two approaches for the fallen person detector. The first approach can differentiate between fallen persons and persons doing ordinary activity in a single stage, while the second is a two-staged approach. The network weights were obtained using a fine-tuning process by retraining with our own extended Fall Person Dataset (E-FPDS), which we release as a benchmark for other RGB vision-based approaches. Quantitative evaluations confirm that the detector performs robustly in detecting fallen persons in different situations. The results also show a recall of 98.97% in our test set.

A Generative Adversarial Network-Based Fault Detection Approach for Photovoltaic Panel

Photovoltaic (PV) panels are widely adopted and set up on residential rooftops and photovoltaic power plants. However, long-term exposure to ultraviolet rays, high temperature and humid environments accelerates the oxidation of PV panels, which finally results in functional failure. The traditional fault detection approach for photovoltaic panels mainly relies on manual inspection, which is inefficient. Lately, machine vision-based approaches for fault detection have emerged, but lack of negative samples usually results in low accuracy and hinders the wide adoption of machine vision-based approaches. To address this issue, we proposed a semi-supervised anomaly detection model based on the generative adversarial network. The proposed model uses the generator network to learn the data distribution of the normal PV panel dataset during training. When abnormal PV panel data are put into the model in the test phase, the reconstructed image generated by the model does not equal the input image. Since the abnormal PV panel data do not obey the data distribution learned by the generator, the difference between the original image and its reconstructed image exceeds the given threshold. So, the model can filter out the fault PV panel by checking the error value between the original image and its reconstructed image. The model adopts Gradient Centralization and SmoothL1 loss function to improve its generalization performance. Meanwhile, we use the convolutional block attention module (CBAM) to make the model pay more attention to the defective area and greatly improve the performance of the model. In this paper, the photovoltaic panels dataset is collected from a PV power plant located in Zhejiang, China. We compare the proposed approach with state-of-the-art semi-supervised and unsupervised approaches (i.e., AnoGAN (Anomaly Detection with Generative Adversarial Networks), Zhao’s method, GANomaly, and f-AnoGAN), and the result indicates that the Area Under Curve (AUC) increases by 0.06, 0.052, 0.041 and 0.035, respectively, significantly improving the accuracy of photovoltaic panel fault detection.

Vision sensor based residual vibration suppression strategy of non-deformable object for robot-assisted assembly operation with gripper flexibility

PurposeIndustrial robots are extensively deployed to perform repetitive and simple tasks at high speed to reduce production time and improve productivity. In most cases, a compliant gripper is used for assembly tasks such as peg-in-hole assembly. A compliant mechanism in the gripper introduces flexibility that may cause oscillation in the grasped object. Such a flexible gripper–object system can be considered as an under-actuated object held by the gripper and the oscillations can be attributed to transient disturbance of the robot itself. The commercially available robots do not have a control mechanism to reduce such induced vibration. Thus, this paper aims to propose a contactless vision-based approach for vibration suppression which uses a predictive vibrational amplitude error-based second-stage controller.Design/methodology/approachThe proposed predictive vibrational amplitude error-based second-stage controller is a real-time vibration control strategy that uses predicted error to estimate the second-stage controller output. Based on controller output, input trajectories were estimated for the internal controller of the robot. The control strategy efficiently handles the system delay to execute the control input trajectories when the oscillating object is at an extreme position.FindingsThe present controller works along with the internal controller of the robot without any interruption to suppress the residual vibration of the object. To demonstrate the robustness of the proposed controller, experimental implementation on Asea Brown Boveri make industrial robot (IRB) 1410 robot with a low frame rate camera has been carried out. In this experiment, two objects have been considered that have a low (<2.38 Hz) and high (>2.38 Hz) natural frequency. The proposed controller can suppress 95% of vibration amplitude in less than 3 s and reduce the stability time by 90% for a peg-in-hole assembly task.Originality/valueThe present vibration control strategy uses a camera with a low frame rate (25 fps) and the delays are handled intelligently to favour suppression of high-frequency vibration. The mathematical model and the second-stage controller implemented suppress vibration without modifying the robot dynamical model and the internal controller.

A radiological image analysis framework for early screening of the COVID-19 infection: A computer vision-based approach

Due to the absence of any specialized drugs, the novel coronavirus disease 2019 or COVID-19 is one of the biggest threats to mankind Although the RT-PCR test is the gold standard to confirm the presence of this virus, some radiological investigations find some important features from the CT scans of the chest region, which are helpful to identify the suspected COVID-19 patients. This article proposes a novel fuzzy superpixel-based unsupervised clustering approach that can be useful to automatically process the CT scan images without any manual annotation and helpful in the easy interpretation. The proposed approach is based on artificial cell swarm optimization and will be known as the SUFACSO (SUperpixel based Fuzzy Artificial Cell Swarm Optimization) and implemented in the Matlab environment. The proposed approach uses a novel superpixel computation method which is helpful to effectively represent the pixel intensity information which is beneficial for the optimization process. Superpixels are further clustered using the proposed fuzzy artificial cell swarm optimization approach. So, a twofold contribution can be observed in this work which is helpful to quickly diagnose the patients in an unsupervised manner so that, the suspected persons can be isolated at an early phase to combat the spread of the COVID-19 virus and it is the major clinical impact of this work. Both qualitative and quantitative experimental results show the effectiveness of the proposed approach and also establish it as an effective computer-aided tool to fight against the COVID-19 virus. Four well-known cluster validity measures Davies–Bouldin, Dunn, Xie–Beni, and β index are used to quantify the segmented results and it is observed that the proposed approach not only performs well but also outperforms some of the standard approaches. On average, the proposed approach achieves 1.709792, 1.473037, 1.752433, 1.709912 values of the Xie–Beni index for 3, 5,7, and 9 clusters respectively and these values are significantly lesser compared to the other state-of-the-art approaches. The general direction of this research is worthwhile pursuing leading, eventually, to a contribution to the community.

Hybrid offline programming method for robotic welding systems

• Current automatic programming approaches for robotic manufacturing systems have been reviewed. • A hybrid offline programming method that combines CAD-based and vision-based offline approaches is proposed. • The vision-based and CAD-based activities of the proposed programming method with the support techniques are presented. • Vision & CAD interactive activities are proposed to identify the workplace, detect the deviation, and compensated the path. Offline programming is an intuitive and automatic programming generation technique that does not use real robotic systems, thus greatly decreasing the downtime required for system programming, and resulting in enormous savings in terms of labor costs. Currently, offline programming can be generally categorized into computer-aided-design-based (CAD-based) and vision-based approaches; these two types of offline programming approaches have been widely applied in robotic welding systems. However, owing to the highly complex and diverse workpieces needed in the shipbuilding industry, neither of the aforementioned offline programming approaches can fully support the automatic generation of welding programs. In this paper, a hybrid offline programming method systematically combining CAD-based, vision-based, and vision & CAD interactive activities is proposed to overcome the limitations of current automatic program generation methods for robotic welding systems. In the vision-based activities, the positions of the workpieces are obtained by using geometrical features gathered from the workpieces’ images, whereas in the CAD-based activities, the welding tasks are assigned to different mobile components of the welding torch; then, their welding paths are planned according to the workpieces’ CAD models. The vision & CAD interactive activities enable the mapping between the point cloud of a workpiece and its CAD model, so that the deviations caused by assembly errors can be detected and path compensation data can be determined. The effectiveness of the proposed hybrid offline programming method is demonstrated by integrating it into a subassembly welding robotic system. The experimental results indicate that this method can significantly improve the efficiency, accuracy, and flexibility of the robotic welding system.

3D-RPP: a novel 3D vision-based Pose Perception Approach for Industrial Robots

Enhanced by artificial intelligence, industrial robots are becoming more powerful, gaining a large variety of applications in intelligent factories. Pose perception, the aim of which is to obtain the joint coordinates of the robot in the camera coordinate system, has been proposed as a promising technology for multi-robot intelligent interaction. A large number of efforts have been made on studying robot pose perception. However, most of existing studies estimate the position of robot joints in 3D space by just using 2D color images, and perform pose perception by estimating the position of joint key points in the image, which could become infeasible in scenarios with various camera views and background environments. To address the issue, considering the information about the robot itself as a prior knowledge, we propose a novel approach for robot pose perception, named 3D-RPP. We adopt a 3D visual point cloud to estimate the rigid transformation of the camera coordinate system with respect to the robot base coordinate system, which effectively improves the accuracy of the obtained robot joint position in the camera coordinate system. We conduct extensive experiments on ROKAE xMate3 robot to investigate the performance of 3D-RPP, and the experimental results show that 3D-RPP could solve the pose perception problem well.

Recognizing workers' construction activities on a reinforcement processing area through the position relationship of objects detected by faster R-CNN

PurposeRecognizing every worker's working status instead of only describing the existing construction activities in static images or videos as most computer vision-based approaches do; identifying workers and their activities simultaneously; establishing a connection between workers and their behaviors.Design/methodology/approachTaking a reinforcement processing area as a research case, a new method for recognizing each different worker's activity through the position relationship of objects detected by Faster R-CNN is proposed. Firstly, based on four workers and four kinds of high-frequency activities, a Faster R-CNN model is trained. Then, by inputting the video into the model, with the coordinate of the boxes at each moment, the status of each worker can be judged.FindingsThe Faster R-CNN detector shows a satisfying performance with an mAP of 0.9654; with the detected boxes, a connection between the workers and activities is established; Through this connection, the average accuracy of activity recognition reached 0.92; with the proposed method, the labor consumption of each worker can be viewed more intuitively on the visualization graphics.Originality/valueWith this proposed method, the visualization graphics generated will help managers to evaluate the labor consumption of each worker more intuitively. Furthermore, human resources can be allocated more efficiently according to the information obtained. It is especially suitable for some small construction scenarios, in which the recognition model can work for a long time after it is established. This is potentially beneficial for the healthy operation of the entire project, and can also have a positive indirect impact on structural health and safety.

A Computer Vision-Based Approach for Tick Identification Using Deep Learning Models.

Simple SummaryTicks are ectoparasites of humans, livestock, and wild animals and, as such, they are a nuisance, as well as vectors for disease transmission. Since the risk of tick-borne disease varies with the tick species, tick identification is vitally important in assessing threats. Standard taxonomic approaches are time-consuming and require skilled microscopy. Computer vision may provide a tenable solution to this problem. The emerging field of computer vision has many practical applications already, such as medical image analyses, facial recognition, and object detection. This tool may also help with the identification of ticks. To train a computer vision model, a substantial number of images are required. In the present study, tick images were obtained from a tick passive surveillance program that receives ticks from public individuals, partnering agencies, or veterinary clinics. We developed a computer vision method to identify common tick species and our results indicate that this tool could provide accurate, affordable, and real-time solutions for discriminating tick species. It provides an alternative to the present tick identification strategies.A wide range of pathogens, such as bacteria, viruses, and parasites can be transmitted by ticks and can cause diseases, such as Lyme disease, anaplasmosis, or Rocky Mountain spotted fever. Landscape and climate changes are driving the geographic range expansion of important tick species. The morphological identification of ticks is critical for the assessment of disease risk; however, this process is time-consuming, costly, and requires qualified taxonomic specialists. To address this issue, we constructed a tick identification tool that can differentiate the most encountered human-biting ticks, Amblyomma americanum, Dermacentor variabilis, and Ixodes scapularis, by implementing artificial intelligence methods with deep learning algorithms. Many convolutional neural network (CNN) models (such as VGG, ResNet, or Inception) have been used for image recognition purposes but it is still a very limited application in the use of tick identification. Here, we describe the modified CNN-based models which were trained using a large-scale molecularly verified dataset to identify tick species. The best CNN model achieved a 99.5% accuracy on the test set. These results demonstrate that a computer vision system is a potential alternative tool to help in prescreening ticks for identification, an earlier diagnosis of disease risk, and, as such, could be a valuable resource for health professionals.

A Comparative Analysis Using Silhouette Extraction Methods for Dynamic Objects in Monocular Vision

Moving or dynamic object analysis continues to be an increasingly active research field in computer vision with many types of research investigating different methods for motion tracking, object recognition, pose estimation, or motion evaluation (e.g. in sports sciences). Many techniques are available to measure the forces and motion of the people, such as force plates to measure ground reaction forces for a jump or running sports. In training and commercial solution, the detailed motion of athlete's available motion capture devices based on optical markers on the athlete's body and multiple calibrated fixed cameras around the sides of the capture volume can be used. In some situations, it is not practical to attach any kind of marker or transducer to the athletes or the existing machinery are being used, while it is required by a pure vision-based approach to use the natural appearance of the person or object. When a sporting event is taking place, there are opportunities for computer vision to help the referee and other personnel involved in the sports to keep track of incidents occurring, which may provide full coverage and analysis in details of the event for sports viewers. The research aims at using computer vision methods, specially designed for monocular recording, for measuring sports activities, such as high jump, wide jump, or running. Just for indicating the complexity of the project: a single camera needs to understand the height at a particular distance using silhouette extraction. Moving object analysis benefits from silhouette extraction and this has been applied to many domains including sports activities. This paper comparatively discusses two significant techniques to extract silhouettes of a moving object (a jumping person) in monocular video data in different scenarios. The results show that the performance of silhouette extraction varies in dependency on the quality of used video data.

Out-of-plane modal property extraction based on multi-level image pyramid reconstruction using stereophotogrammetry

Understanding the dynamic behavior of as-built structures is important, because structural condition can often be inferred from changes in its dynamic response. To this end, modal properties such as natural frequencies, damping ratios, and mode shapes can be extracted from acceleration measurements. Associated installations require sensors, cabling, and data acquisition systems that can be expensive and time consuming. Alternatively, computer vision-based approaches have been proposed that offer non-contact measurements, as well as substantial cost savings. However, such approaches usually focus only on the structural component responses that are in the plane of the image; researchers seldom consider the structural response that is out-of-plane (i.e., perpendicular to the image plane). For complex structural component that possess asymmetries, the out-of-plane behavior is critical to understanding their dynamic response. Determination of dynamic response of the structure requires that multiple cameras are used and can be computationally demanding. In this study, imagery from a commercially available stereo camera is combined with a multi-level image pyramid approach and operational modal analysis (OMA) to extract out-of-plane modal properties. The measurements are first padded to reduce the distortion effect caused by the image pyramid. The responses are then compressed and decomposed into sub-bands using an image pyramid decomposition; followed by extraction of the modal properties using the OMA approach known as frequency-domain stochastic subspace identification. The advantages and limitations of the proposed approach are illustrated numerically using a physics-based graphics model (PBGM) of a continuous beam. Subsequently, experimental validation is conducted for a 3-story model building using the Intel® RealSenseTM D415 depth camera. Modal properties are shown to be determined quickly, with high accuracy and noise robustness. Moreover, detailed near-continuous mode shapes are obtained using multi-level image pyramid reconstruction. These results demonstrate that the proposed approach can give an accurate picture of the dynamic characteristics of a structure, offering the potential for effective long-term structural health monitoring for important structural component.

Contour Tracking Control for Mobile Robots applicable to Large-scale Assembly and Additive Manufacturing in Construction

In the construction industry, as well as during the assembly of large-scale components, the required workspaces usually cannot be served by a stationary robot. Instead, mobile robots are used to increase the accessible space. Here, the problem arises that the accuracy of such systems is not sufficient to meet the tolerance requirements of the components to be produced. Furthermore, there is an additional difficulty in the trajectory planning process since the exact dimensions of the pre-manufactured parts are unknown. Hence, existing static planning methods cannot be exerted on every application. Recent approaches present dynamic planning algorithms based on specific component characteristics. For example, the latest methods follow the contour by a force-controlled motion or detect features with a camera. However, in several applications such as welding or additive manufacturing in construction, no contact force is generated that could be controlled. Vision-based approaches are generally restricted by varying materials and lighting conditions, often found in large-scale construction. For these reasons, we propose a more robust approach without measuring contact forces, which, for example, applies to large-scale additive manufacturing. We based our algorithm on a high-precision 2D line laser, capable of detecting different feature contours regardless of material or lightning. The laser is mounted to the robot’s end-effector and provides a depth profile of the component’s surface. From this depth data, we determine the target contour and control the manipulator to follow it. Simultaneously we vary the robot’s speed to adjust the feed rate depending on the contour’s shape, maintaining a constant material application rate. As a proof of concept, we apply the algorithm to the additive manufacturing of two-layer linear structures made from spray PU foam. When making these structures, each layer must be positioned precisely on the previous layer to obtain a straight wall and prevent elastic buckling or plastic collapse. Initial experiments show improved layer alignment within 10 % of the layer width, as well as better layer height consistency and process reliability.

Vision Based Real Time Monitoring System for Elderly Fall Event Detection Using Deep Learning

Human fall detection plays a vital part in the design of sensor based alarming system, aid physical therapists not only to lessen after fall effect and also to save human life. Accurate and timely identification can offer quick medical services to the injured people and prevent from serious consequences. Several vision-based approaches have been developed by the placement of cameras in diverse everyday environments. At present times, deep learning (DL) models particularly convolutional neural networks (CNNs) have gained much importance in the fall detection tasks. With this motivation, this paper presents a new vision based elderly fall event detection using deep learning (VEFED-DL) model. The proposed VEFED-DL model involves different stages of operations namely preprocessing, feature extraction, classification, and parameter optimization. Primarily, the digital video camera is used to capture the RGB color images and the video is extracted into a set of frames. For improving the image quality and eliminate noise, the frames are processed in three levels namely resizing, augmentation, and min–max based normalization. Besides, MobileNet model is applied as a feature extractor to derive the spatial features that exist in the preprocessed frames. In addition, the extracted spatial features are then fed into the gated recurrent unit (GRU) to extract the temporal dependencies of the human movements. Finally, a group teaching optimization algorithm (GTOA) with stacked autoencoder (SAE) is used as a binary classification model to determine the existence of fall or non-fall events. The GTOA is employed for the parameter optimization of the SAE model in such a way that the detection performance can be enhanced. In order to assess the fall detection performance of the presented VEFED-DL model, a set of simulations take place on the UR fall detection dataset and multiple cameras fall dataset. The experimental outcomes highlighted the superior performance of the presented method over the recent methods.

Coordination Control of the Automated Fiber Placement System Using Vision-Based Leader-Follower Approach

Coordination Control of the Automated Fiber Placement System Using Vision-Based Leader-Follower Approach