Crowd density estimation using deep learning for Hajj pilgrimage video analytics

Background: This paper focuses on advances in crowd control study with an emphasis on high-density crowds, particularly Hajj crowds. Video analysis and visual surveillance have been of increasing importance in order to enhance the safety and security of pilgrimages in Makkah, Saudi Arabia. Hajj is considered to be a particularly distinctive event, with hundreds of thousands of people gathering in a small space, which does not allow a precise analysis of video footage using advanced video and computer vision algorithms. This research proposes an algorithm based on a Convolutional Neural Networks model specifically for Hajj applications. Additionally, the work introduces a system for counting and then estimating the crowd density. Methods: The model adopts an architecture which detects each person in the crowd, spots head location with a bounding box and does the counting in our own novel dataset (HAJJ-Crowd). Results: Our algorithm outperforms the state-of-the-art method, and attains a remarkable Mean Absolute Error result of 200 (average of 82.0 improvement) and Mean Square Error of 240 (average of 135.54 improvement). Conclusions: In our new HAJJ-Crowd dataset for evaluation and testing, we have a density map and prediction results of some standard methods.

Trends in AI Research for the Visual Surveillance of Populations

Progress in artificial intelligence has led to growing concern about the capabilities of AI-powered surveillance systems. This data brief uses bibliometric analysis to chart recent trends in visual surveillance research — what share of overall computer vision research it comprises, which countries are leading the way, and how things have varied over time.

Oil Spill Detection Using Fluorometric Sensors: Laboratory Validation and Implementation to a FerryBox and a Moored SmartBuoy

A large part of oil spills happen near busy marine fairways. Presently, oil spill detection and monitoring are mostly done with satellite remote sensing algorithms, or with remote sensors or visual surveillance from aerial vehicles or ships. These techniques have their drawbacks and limitations. We evaluated the feasibility of using fluorometric sensors in flow-through systems for real-time detection of oil spills. The sensors were capable of detecting diesel oil for at least 20 days in laboratory conditions, but the presence of CDOM, turbidity and algae-derived substances substantially affected the detection capabilities. Algae extract was observed to have the strongest effect on the fluorescence signal, enhancing the signal in all combinations of sensors and solutions. The sensors were then integrated to a FerryBox system and a moored SmartBuoy. The field tests support the results of the laboratory experiments, namely that the primary source of the measured variation was the presence of interference compounds. The 2 month experiments data did not reveal peaks indicative of oil spills. Both autonomous systems worked well, providing real-time data. The main uncertainty is how the sensors' calibration and specificity to oil, and the measurement depth, affects oil detection. We recommend exploring mathematical approaches and more advanced sensors to correct for natural interferences.

Crowd density estimation using deep learning for Hajj pilgrimage video analytics

Background: This paper focuses on advances in crowd control study with an emphasis on high-density crowds, particularly Hajj crowds. Video analysis and visual surveillance have been of increasing importance in order to enhance the safety and security of pilgrimages in Makkah, Saudi Arabia. Hajj is considered to be a particularly distinctive event, with hundreds of thousands of people gathering in a small space, which does not allow a precise analysis of video footage using advanced video and computer vision algorithms. This paper aims to propose an algorithm based on a Convolutional Neural Networks model specifically for Hajj applications. Additionally, the work introduces a system for counting and then estimating the crowd density. Methods: The model adopts an architecture which detects each person in the crowd, spots head location with a bounding box and does the counting in our own novel dataset (HAJJ-Crowd). Results: Our algorithm outperforms the state-of-the-art method, and attains a remarkable Mean Absolute Error result of 200 (average of 82.0 improvement) and Mean Square Error of 240 (average of 135.54 improvement). Conclusions: In our new HAJJ-Crowd dataset for evaluation and testing, we have a density map and prediction results of some standard methods.

Gender Classification Using Proposed CNN-Based Model and Ant Colony Optimization

Pedestrian gender classification is one of the key assignments of pedestrian study, and it finds practical applications in content-based image retrieval, population statistics, human–computer interaction, health care, multimedia retrieval systems, demographic collection, and visual surveillance. In this research work, gender classification was carried out using a deep learning approach. A new 64-layer architecture named 4-BSMAB derived from deep AlexNet is proposed. The proposed model was trained on CIFAR-100 dataset utilizing SoftMax classifier. Then, features were obtained from applied datasets with this pre-trained model. The obtained feature set was optimized with ant colony system (ACS) optimization technique. Various classifiers of SVM and KNN were used to perform gender classification utilizing the optimized feature set. Comprehensive experimentation was performed on gender classification datasets, and proposed model produced better results than the existing methods. The suggested model attained highest accuracy, i.e., 85.4%, and 92% AUC on MIT dataset, and best classification results, i.e., 93% accuracy and 96% AUC, on PKU-Reid dataset. The outcomes of extensive experiments carried out on existing standard pedestrian datasets demonstrate that the proposed framework outperformed existing pedestrian gender classification methods, and acceptable results prove the proposed model as a robust model.

Detecting Driver Drowsiness with Multi-Sensor Data Fusion Combined with Machine Learning

In this research work, we develop a drowsy driver detection system through the application of visual and radar sensors combined with machine learning. The system concept was derived from the desire to achieve a high level of driver safety through the prevention of potentially fatal accidents involving drowsy drivers. According to the National Highway Traffic Safety Administration, drowsy driving resulted in 50,000 injuries across 91,000 police-reported accidents, and a death toll of nearly 800 in 2017. The objective of this research work is to provide a working prototype of Advanced Driver Assistance Systems that can be installed in present-day vehicles. By integrating two modes of visual surveillance to examine a biometric expression of drowsiness, a camera and a micro-Doppler radar sensor, our system offers high reliability over 95% in the accuracy of its drowsy driver detection capabilities. The camera is used to monitor the driver’s eyes, mouth and head movement and recognize when a discrepancy occurs in the driver's blinking pattern, yawning incidence, and/or head drop, thereby signaling that the driver may be experiencing fatigue or drowsiness. The micro-Doppler sensor allows the driver's head movement to be captured both during the day and at night. Through data fusion and deep learning, the ability to quickly analyze and classify a driver's behavior under various conditions such as lighting, pose-variation, and facial expression in a real-time monitoring system is achieved.

Visual presentation for monitoring layer-wise curing quality in DLP 3D printing

Purpose This paper aims to address the problem of uncertain product quality in digital light processing (DLP) three-dimensional (3D) printing, a scheme is proposed to qualitatively estimate whether a layer is printed with the qualified quality or not cured . Design/methodology/approach A thermochromic pigment whose color fades at 45°C is prepared as the indicator and it is mixed with the resin. A visual surveillance framework is proposed to monitor the visual variation in a period of the entire curing process. The exposure region is divided into 30 × 30 sub-regions; gray-level variation curves (curing curves) in all sub-regions are classified as normal or abnormal and a corresponding printing control strategy is designed to improve the percentage of qualified printed objects. Findings The temperature variation caused by the releasing reaction heat on the exposure surface is consistent in different regions under the homogenized light intensity. The temperature in depth begins to rise at different times. The temperature in the regions near the light source rises earlier, and that far from the light source rises later. Thus, the color of resin mixed with the thermochromic pigment fades gradually over a period of the entire solidification process. The color variation in the regions with defects of bubbles, insufficient material filling, etc., is much different from that in the normal curing regions. Originality/value A temperature-sensitive organic chromatic chemical pigment is prepared to present the visual variation over a period of the entire curing process. A novel 3D printing scheme with visual surveillance is proposed to monitor the layer-wise curing quality and to timely stop the possible unqualified printing resulted from bubbles, insufficient material filling, etc.