Using Artificial Intelligence for CCTV Video Analytics to Improve Security: Ipsotek

The 3-dimensional video motion detection system (3DVMD) is a stand-alone sensor system developed at Sandia National Laboratories. It operates on a single PC platform and uses multiple cameras to monitor activity in predefined, 3-dimensional volumes of space. Detected activity is displayed on the PC monitor as an overlay to real-time video imagery. The basic sensor can provide switch closures for an interface to an Alarm Communications and Display (AC&D) system. High-level detection information relating to the size, location, and direction of motion of moving objects can be provided for assessment purposes. Using this high-level information, 3DVMD can be used to monitor individual items as well as track people within the monitored area. It can be used for two-man rule verification, process control, and the detection of anomalies during evacuation procedures. The 3DVMD system can be set up as an interior or exterior sensor. There are several benefits to the 3DVMD technology. 3DVMD is a passive security sensor. It provides high-level information (e.g., location, size, direction of motion) that can be used as an assessment aid in CCTV video, where it is often difficult to even see an intruder. Its detection volume is user-definable and remains static until changed by the user. This reduces vulnerabilities introduced by the variable detection envelope of other sensors. It has a reduced nuisance alarm rate as compared to traditional 2-dimensional video motion detection sensors. The technology also has built-in redundancy. If a camera fails, the system can adjust to operate without that camera, down to a minimum of two cameras. 3DVMD also has a self-assessment capability. Because CCTV sensors are an integral component of the 3DVMD system, the failure or tampering of any CCTV sensor is immediately identified. 3DVMD technology is ready to be fielded either as a stand-alone security system or as an integrated security system that provides switch closures to an existing AC&D system. Minimal software development is required to customize the technology for specific applications. Sandia is currently in the technology transfer process for the baseline 3DVMD technology.

With the development of deep learning, artificial intelligence applications and services have boomed in the recent years, including recommendation systems, personal assistant and video analytics. Similar to other services in the edge computing environment, artificial intelligence computing tasks are pushed to the network edge. In this paper, we consider the multi-user edge-assisted video analytics task offloading (MEVAO) problem, where users have video analytics tasks with various accuracy requirements. All users independently choose their accuracy decisions, satisfying the accuracy requirement, and offload the video data to the edge server. With the utility function designed based on the features of video analytics, we model MEVAO as a game theory problem and achieve the Nash equilibrium. For the flexibility of making accuracy decisions under different circumstances, a deep reinforcement learning approach is applied to our problem. Our proposed design has much better performance compared with some other approaches in the extensive simulations.

Recent advances in machine learning and deep learning have enabled many existing applications in smart cities, autonomous systems, and wearable devices. These applications often demand scalable real-time cognitive intelligence and on-the-spot decision making. Current computer systems have been customized for a cloud computing paradigm which often does not meet latency constraints and scalability requirements. To address the limitations of the cloud computing paradigm, the general trend is toward shifting the computation next to data producers at the edge. However, the edge computing paradigm is in the very early stages. Many system-level aspects of edge computing, including algorithms mapping and partitioning across edge computing resources (edge server, and edge nodes) are unknown. New research is required to understand and quantify design dimensions for edge computing.This paper presents a novel edge computing infrastructure for distributed real-time video analytics. This paper presents a holistic solution for co-designing application and edge infrastructure, including edge nodes and edge servers, to enable scalable real-time Artificial Intelligence (AI)/Deep Learning (DL) video analytics across many cameras. For experimental results and evaluation, we focus on the case study of object re-identification across many cameras, which is composed of object detection/classification (TinyYOLOv3), feature extraction, local re-identification, and global re-identification kernels. We evaluate the edge system under three different task mapping and resource allocation configurations. The results present that with the edge nodes (video cameras) more than 32, the only scalable solution is to perform detection/classification (TinyYOLOv3), feature extraction, local re-identification on the edge nodes next to cameras, and execute global re-identification on edge server.

Esophageal cancer surgery is evolving from technical standardization to a paradigm of personalized, strategy‐oriented care. Robotic‐assisted techniques and enhanced perioperative protocols have improved safety, but the field is increasingly shaped by three forces: integration of immune checkpoint inhibitors (ICIs), population aging, and the rise of intelligent technologies such as artificial intelligence (AI) and extended reality. Adjuvant nivolumab after neoadjuvant chemoradiotherapy (nCRT) remains the standard for residual disease, while other regimens such as durvalumab or dual checkpoint blockade have not demonstrated consistent survival benefit. Neoadjuvant ICI strategies, particularly camrelizumab plus chemotherapy in esophageal squamous cell cancer (ESCC), achieve high pathological response rates without increasing surgical morbidity, and pooled analyses confirm their feasibility. Immune‐related adverse events (irAEs) occur in approximately 20%–35% of patients but are usually manageable. Perioperative outcomes remain comparable to conventional regimens. Elderly and frail patients require individualized optimization. The integration of minimally invasive techniques, assessment of sarcopenia and nutritional risk, and adjustment of chemotherapy intensity have contributed to improved outcomes. Moreover, salvage surgery—once prohibitive—now offers meaningful long‐term survival when performed in high‐volume centers with specialized expertise. AI and machine learning are transforming risk stratification, intraoperative guidance, and surgical training. AI‐assisted video analytics and VR/AR simulators enhance skill acquisition, credentialing, and standardization. Future progress will depend on multicenter validation, prospective registries, and integration of oncologic, physiologic, and technological variables. Ultimately, the future of esophageal cancer surgery will be defined not only by technical precision but also by the surgeon's ability to leverage data‐driven innovation for personalized care.

Simple SummaryOur research team has developed an automated computer system that uses convolutional neural networks (CNNs) to monitor and analyse the sleep patterns of dogs. Traditional methods of recording animal behaviour, such as direct observations (of sleep) of either live behaviour or recorded behaviour, can be time-consuming and error-prone, making it difficult to replicate studies. Sleep may be a crucial indicator of an animal’s well-being, but it has been overlooked in animal welfare research due to the time-consuming nature of measuring sleep. Compared to direct behavioural observations from the same videos, our system achieved an 89% similarity score in automatically detecting and quantifying sleep duration and fragmentation in dogs. Although there were no significant differences in the time percentage of sleep observed, the system recorded more total sleep time than human observers making direct observations on the same data sources. The automated system used could become a valuable tool for animal behaviour and welfare research.Although direct behavioural observations are widely used, they are time-consuming, prone to error, require knowledge of the observed species, and depend on intra/inter-observer consistency. As a result, they pose challenges to the reliability and repeatability of studies. Automated video analysis is becoming popular for behavioural observations. Sleep is a biological metric that has the potential to become a reliable broad-spectrum metric that can indicate the quality of life and understanding sleep patterns can contribute to identifying and addressing potential welfare concerns, such as stress, discomfort, or health issues, thus promoting the overall welfare of animals; however, due to the laborious process of quantifying sleep patterns, it has been overlooked in animal welfare research. This study presents a system comparing convolutional neural networks (CNNs) with direct behavioural observation methods for the same data to detect and quantify dogs’ sleeping patterns. A total of 13,688 videos were used to develop and train the model to quantify sleep duration and sleep fragmentation in dogs. To evaluate its similarity to the direct behavioural observations made by a single human observer, 6000 previously unseen frames were used. The system successfully classified 5430 frames, scoring a similarity rate of 89% when compared to the manually recorded observations. There was no significant difference in the percentage of time observed between the system and the human observer (p > 0.05). However, a significant difference was found in total sleep time recorded, where the automated system captured more hours than the observer (p < 0.05). This highlights the potential of using a CNN-based system to study animal welfare and behaviour research.

With the rapid development of the Internet and the rapid improvement of smartphone technology, a large number of short videos are shared through social platforms every minute, so video content analysis is currently a very important and popular job in machine learning and artificial intelligence. In recent years, in the field of domestic Internet content, platforms such as Toutiao, Netease, Yidian Information, and Baijiahao have sprung up. These content distribution platforms utilize multi-channel content sources, and the amount of content has exploded, of which short video content accounts for a large proportion. The rich short video content has changed the behavior of users and provided users with a more convenient way to socialize. Short video has a higher dimension of information presentation, and it is easier for people to accept and spread. At present, a large amount of short video information has appeared on the Internet. These short video information have brought serious information overload problems to users, and also brought huge challenges to short video operators and video editors. This paper adopts the method of multimodal fusion, which has high accuracy in identifying emotional states such as happiness, sadness, anger, and disgust in the sample.

The escalation of security challenges in the context of digital transformation highlights the need for a systematic review of current practices, risks and the potential for implementing artificial intelligence in law enforcement activities. The aim of this study was to summarise scientific approaches to the application of artificial intelligence in law enforcement, focusing on the stages of its development, key areas of research and insufficiently studied aspects. The use of methods of analysis and synthesis of scientific sources, content analysis, comparative analysis, and classification of existing approaches made it possible to assess the current state of scientific research on trends, challenges, and prospects for the use of artificial intelligence. It has been established that scientific interest in the application of artificial intelligence in law enforcement has increased significantly over the last decade. The rapid development of artificial intelligence technologies has opened up new opportunities for the automation of analytical and operational functions, prompting scientists to study the possibilities and threats of artificial intelligence. Researchers focus primarily on areas such as video analytics, crime prediction, image recognition, and big data processing. At the same time, there is a lack of in- depth interdisciplinary research that takes into account the ethical, legal, and social implications of using such technologies. A disparity in approaches to risk classification and standardisation of implementation practices has been noted. The need for the formalisation of research has been demonstrated, which will contribute to the balanced development of artificial intelligence in law enforcement activities, taking into account security, legal, and humanitarian factors. The results obtained can be used by heads of law enforcement agencies, analytical units, and digital transformation specialists to determine priority development directions and consider potential risks

Video analytics powered by artificial intelligence shows high promise in making our society smarter. Harnessing large amounts of video data, however, requires the development of processing systems demonstrating high performance and high efficiency. To this end, this work has contributed to a video analytics system powered by artificial intelligence for object detection and recognition. Rather than streaming all the video frames to the cloud, the system analyzes images on-camera and only returns those of interest to the cloud. This edge analytics research-grade software is available, but it lacks a simple web interface for general use by scientists, engineers, and other experts. To make the system versatile and user-friendly, a web interface was developed for receiving user queries (including parameters such as camera geolocations, video time span, and object class of interest) and presenting the query results. This web interface will make our video analytics system more accessible to domain experts (those in law enforcement, health care, environmental monitoring, etc.).

The fusion of Artificial Intelligence (AI) and Machine Learning (ML) with the existing Indian Railways CCTV network provides revolutionary potential to improve safety, security, and operational efficiency. This research showcases the creation of an AI-based railway surveillance system that is meant to solve issues related to controlling congested stations, deterring crimes, and ensuring efficient workforce monitoring. Through the use of sophisticated video analytics, deep learning algorithms, and computer vision technology, the system facilitates real-time anomaly detection, prevention of unauthorized access, and predictive analytics to optimize crowd management. In addition, AI-powered facial recognition and behavioral analysis models optimize crime prevention by detecting suspicious behavior and known criminals. Modules for work monitoring guarantee compliance with safety procedures among staff and streamline station operations. The research also examines the technological and ethical aspects of having such a system, for instance, high-resolution cameras, edge computing, data privacy, and regulatory issues. The outcome identifies that AI-powered surveillance has the potential to significantly improve passenger safety, utilization of resources, and efficiency in operations, making Indian Railways ready to embrace a safer, brighter future Key Words: Artificial Intelligence, Machine Learning, Railway Surveillance, Video Analytics, Crime Prevention, Operational Efficiency.

The recognition of human activities using vision based methods has become a key area of interest in video analytics research. In the past years, several sophisticated deep learning (DL) algorithms have surfaced, designed to recognize complex human actions depicted in video streams. These approaches have exhibited remarkable competence in tasks related to video analytics. This paper explores the field in greater depth by centering its focus on Sign Language (SL) as a unique and complex form of human activity. SL plays is pivotal in facilitating communication for people who are deaf or hard of hearing, enabling them to interact and engage. The main idea is to recognize Arabic Sign Language especially Tunisian accent, and subsequently translate it into spoken language utilizing artificial intelligence (AI) methods. The goal of this work is to facilitate communication between individuals who are deaf and those who are not. So, this paper presents the comparative results of four deep-learning models: Long Short-Term Memory (LSTM), Bidirectional Long Short-Term Memory (Bi-LSTM), Gated Reccurent Unit (GRU), and Bidirectional Gated Reccurent Unit (Bi-GRU), aiming to improve the recognition accuracy of dynamic signs. The results indicate that the BiLSTM outperforms other variants, showcasing an accuracy of 98%.

Recently, Coronavirus Disease (COVID-19) has spread rapidly across the world and thus social distancing and face mask has become one of the mandatory preventive measures. The Artificial Intelligence community has been focusing on developments for monitoring social distancing and identifying face masks, which have become the hotspot and made the headlines. A powerful approach for fast and more accurate detection and monitoring would greatly enhance the sole purpose, save the lives of many and also alleviate the burden of doing it manually to some extent. You Only Look Once(Yolov3) and improved Single Shot multi-box Detector (SSD) algorithm has been deployed for an improved object detection model to lend a helping hand in the fight against this deadly disease. The proposed framework utilizes Single Shot multi-box Detector (SSD) with MobileNetV2, to enhance the feature extraction ability of the object detection model. After taking into account the significance of developing an accurate model and also the limitations of existing models, herein we have proposed a framework based on deep learning, which would automate and simplify the task of monitoring social distancing and face mask through intelligent video analytics.

Background The advent of artificial intelligence (AI) in surgical training presents a transformative opportunity to enhance the assessment and education of surgical fellows. This study investigates the application of AI in evaluating the performance of surgeons by analyzing videos of laparoscopic sleeve gastrectomy procedures uploaded to a specialized platform, Surgeryview.ai. Methods Three surgeons contributed to this study: two were professor bariatric surgeons and one was a fellow bariatric surgeon in training. A total of 466 videos of gastric sleeve surgeries were analyzed. Of these, 413 videos were performed by the professor surgeons (Surgeon 1: 180, Surgeon 2: 233) between September 2022 and May 2024, and 53 surgeries were performed by the fellow surgeon in training over a period of 10 months (August 2023–May 2024). The analyzed videos were chosen randomly, and any surgeries Involving an additional procedure to the gastric sleeve (for example hiatal hernia repair) were excluded. Patient data, including gender, age, and initial BMI, were recorded, ensuring anonymity. Results The AI platform provided comprehensive longitudinal analytics, detailing the time taken to complete each surgical step: access to the abdominal cavity, greater curvature dissection, stapling, specimen removal, oversewing, and leak test. The platform also analyzed the impact of patient demographics on surgery time. Professors showed consistent efficiency, while fellows demonstrated significant improvement over time. Notably, surgery duration was influenced by patient-specific factors such as BMI and age. The fellow bariatric surgeon in training showed significant improvement across all metrics over the trimesters, with overall surgery time reduced by approximately 23.83%, greater curvature dissection (step 2) time by 28.37%, and oversuture of the staple line (step 5) time by 28.74%. When broken down by BMI, age, and gender, indicate that improvements are generally consistent across these variables, with the most significant improvements seen in lower BMI and younger age groups. These advancements indicate that the fellow has substantially closed the performance gap, particularly in the critical surgical steps. Conclusion AI assisted platforms like Surgeryview.ai offer valuable insights into surgical performance, enabling objective assessment and personalized feedback for fellows. This technology can enhance surgical training programs by identifying specific areas for improvement and tracking progress over time, ultimately aiming to standardize surgical education and improve patient outcomes.

Artificial Intelligence (AI) is transforming pediatric orthopedics by harnessing the power of technology to address the unique challenges associated with the growing musculoskeletal system and to adapt to modern healthcare delivery models. This article reviews the literature on recent advancements in AI impacting paediatric orthopedic practices and summarizes our own experience in the development and application of AI in pediatric orthopedics. AI-driven imaging techniques, such as automated ultrasound analysis, have significantly improved screening for developmental dysplasia of the hip (DDH) by enhancing accuracy and accessibility while minimizing radiation exposure. Similarly, machine learning has automated and improved diagnostic precision in complex assessments like bone age determination. In surgical planning, AI-powered robotic systems have optimized spinal instrumentation placement, reduced complications and enhance procedural accuracy. AI-based predictive modeling allows surgeons to develop personalized operative strategies through detailed 3D modeling and simulation, critical for pediatric deformity correction. Remote gait analysis via wearable sensors and smartphone-based video analytics facilitates continuous monitoring, personalized rehabilitation, and early complication detection. Clinical workflow efficiency is improved through AI-enhanced documentation, virtual assistants, and scheduling optimization tools. AI promises substantial benefits for pediatric orthopedic care. Integration allows for more precise diagnosis, treatment personalization, and positive long-term patient outcomes. Ongoing research should focus on clinical adoption strategies and expanding the scope of conditions to which the technology can be applied to maximize the benefit of AI in pediatric orthopedics.

With the development of artificial intelligence technology and intelligent devices, people show great interest in intelligent applications and services, but it is impossible to complete these compute-intensive AI tasks locally, especially video analysis tasks. Edge computing is regarded as an appropriate solution to these problems. In this paper, we study the multi-user multi-server edge-end collaboration video analytics task offloading problem aiming at minimizing the overall delay for each device to finish its task. Each device chooses whether to execute the task locally or to offload the task to an edge server, and which edge server to select. At the theoretical level, we model the joint problem of task offloading and resource allocation as a mixed integer programming problem. We first determine the optimal resource allocation policy with a given task offloading decision profile. Then, task offloading problem is modeled as a congestion game and propose a decentralized mechanism to achieve a Nash equilibrium. Moreover, experimental results demonstrate that the proposed method is efficient and can significantly and steadily improve the system performance, reducing the overall delay by 33.96% on average, compared with other algorithms.

As artificial intelligence continues to evolve, it is increasingly capable of handling a wide range of video analytics tasks with a single large model. One of the key foundation technologies is the Segment Anything Model (SAM), which enables video analytics tasks to be determined on the fly based on user input prompts. However, achieving real-time response in video analytics applications is crucial for user experiences due to the limited communication and computation resources on the edge, especially with SAM, where users may continuously interact by adding or adjusting prompts.