Description:Field of the Invention
The proposed method explains surveillance and security by keeping an eye on and upholding the organization's safety and norms to stop any odd behaviors.
Background of the Invention
The most widely used device to continuously monitor a confined area is closed-circuit television surveillance, which is utilized by many organizations, businesses, homes, workplaces, traffic lights, ATMs, etc. It has had a remarkable evolution in the period of safety and security, improving technologically day by day and becoming essential in the modern world. The goal of CCTV is to give video or picture evidence that may be used to identify the individual who is suspected when some suspicious behavior is observed. With the aid of AI, improvements are incorporated into devices with sophisticated functions in addition to the fundamental functionality.
US8855363B2 also refers to a system in which one embodiment, a method to track people comprises providing first and second images containing a plurality of people, locating faces of people in each image, and generating a facial coefficient vector for each face by extracting from the images coefficients sufficient to locally identify each face, followed by tracking the people within the images, the tracking including comparing the first and second sets of facial coefficient vectors. The approach for tracking people involves utilizing estimated positions in conjunction with the distance between face coefficient vectors.
JP4724125B2 similarly deals with face detection and recognition systems having wide range of uses, including face-based content-based picture retrieval, surveillance systems, and intelligent human computer interfaces. The efficient face attempts to develop detection and face recognition systems, however, have not been successful because the computer and algorithm used in the conventional face detection system and face recognition system could not efficiently process the enormous amount of data and complicated calculations inherent in the face recognition system.
US10757377B2 claims that a surveillance system and its operating mechanism that has a primary sensor module that is operating normally and a secondary sensor module that is in a power-saving mode. Only when the signals produced by the auxiliary sensor module satisfy a surveillance condition is the primary sensor in use. Due to the primary sensor module's increased power remaining in the power-saving mode without being activated, the surveillance system can use less energy. Since the surveillance system employs the procedures of ROI detection, feature extraction, and object recognition at an early stage of a full surveillance process, it is able to handle the surveillance data produced by the auxiliary sensor module.
WO2019127365A1 refers toface recognition technology which immediately acquire the camera through the camera, unlike other biometric recognition methods. The embodiment of the present application provides a method for detecting a live body of a human face. Images are constantly collected, a face is tracked for each frame, and it is determined that the image is a living image if the face tracked for each frame is the same face. It efficiently prevents the fraudulent conduct of stealing images, movies, or other people's faces during the detection of the human face, and determines whether the image is a living image or not by continually tracking each frame of the face in the image. Realize the function of differentiating the actual person from the dummy, and effectively prohibit the behavior of tricking the face recognition system using face images or face videos.
US6594629B1 has been suggested to add visual clues to voice recognition software in an effort to enhance this capability. The majority of efforts in this field, unlike LVCSR efforts, can be considered to be only preliminary because tasks have typically been limited to small vocabulary (e.g., commands, digits) and speaker dependent training or isolated word speech where word boundaries are artificially well defined. This approach has attracted the interest of researchers over the past couple of years.
As mentioned earlier, there must be a mechanism in the information that offers a real-time surveillance system with facial recognition and motion detection technology. The need for AI implementation in various surveillance systems, including video management, optical flow measurement, data management, improved camera capabilities, and security systems, will increase the effectiveness of closed-circuit television systems' support. As a result, it was suggested to analyze the requirements for facial recognition and motion detection in CCTV.
Summary of the Invention
The chore of keeping track of everyone's actions in an educational institutions/confined area is challenging and time-consuming. The supervision of educational institutions with several blocks requires a lot of personnel. It is quite difficult to continuously keep an eye on every nook and cranny and to report every detail to the security department. As a result, the suggested system also uses automatic voice recognition (ASR) to monitor the campus independently. By obtaining the information from the database, the neural network software aids in the identification of a person's face. It categorizes the individual as known or unknown and notifies the security division that the unknown is an invader. The motion detection feature then carefully monitors the unknown person's mobility and warns the security division if any unexpected activities are carried out. If a recognized individual is discovered somewhere other than where they are supposed to be, the details from the database are retrieved and sent to the concerned department. As a result, the suggested system also uses automatic voice recognition (ASR) to monitor the campus independently
Brief Description of Drawings
The invention will be described in detail with reference to the exemplary embodiments shown in the figures wherein:
Figure-1: Flowgorithm representing the work flow of a CCTV in an institution
Figure-2: Diagramatic representation of the working of the invention
Detailed Description of the Invention
In educational institutions, CCTV facial recognition and motion detection are useful for surveillance. Face recognition and motion detection are its two main parts. Face recognition aids in identifying persons who are affiliated with their institutions and identifies intruders who are not listed in the database that contains all the face images of students, faculty, and staff. The person are classified as known if their details are stored in database, if not they are categorised as unknowns.
Face recognition compares the images already present in the database and notifies the operator about the intruder. If a recognized individual is discovered somewhere other than where they are supposed to be, motion detection alerts the handler and lets them know. Motion detection also enables us to determine whether someone is lurking outside the institute at an odd hour. When someone tries to enter the building, security is alerted.
The proposed system provides three outputs namely face detection and identification information and motion detection information. Most traditional and finest algorithms for face detection is Viola-Jones algorithmproposed by Paul Viola and Michael Jones in 2001. It is known for its high detection rate in mostly real time applications. Based on the principle of optical flow, when it comes to motion detection, Lucas-Kanade algorithm is implemented. It works by sampling images at a predetermined frame rate and detecting the movement by observing changes in the pixels.
This process is achieved by:
(i) Image enhancement is employed for modifying an image using specified operations to improve quality based on a particular application goal using histogram and gamma correction techniques. When there is a problem with the image intensity, such as when the image is too dark or light, gamma correction is called for. Histogram equalization is helpful for improving contrast on images.
(ii)Accumulative Differences Images (ADI):Here, motion segmentation is carried out and references from numerous images, not just one, are compared. After comparing the tested images to the reference images, the findings will be tallied up and compared to a predetermined threshold.Threshold contains a typical value that is iterated 'n' times depending on the movement of the object. It is the particular value that governs how an object is moving respectively.
(iii)Local Binary Pattern(LBP) cascade classifier is an algorithm that is better than the traditional Haar cascade algorithm with 3x faster ability. LBP is finest in detecting object in images despites of scale and location of the image.
Key techniques used are:
a) Internal image
b) AdaBoost-based algorithm learning
c) Cascade method for merging classifiers
which process the mechanism and the work flow of the mentioned methods precisely with traditional mathematical reference from various principle algorithms in order to achieve the goal.
By combining these methods with Convolutional Neural Networks (CNN), facial identification is accomplished through picture processing.And coming to motion detection, in contrast to conventional CNNs, which analyze individual frames, 3D CNNs analyze spatiotemporal data by taking several frames into account at once. Within a single network, they may record both geographical and temporal data. The C3D and I3D architectures are frequently used methods for 3D CNN-based motion detection tasks.
Automatic speech recognition (ASR) is a technique that trains a machine to transform spoken words into written text. ASR systems have significantly improved because to machine learning, especially when utilizing natural language processing (NLP) methods.Using sizable datasets with known transcriptions, ASR models are trained. The acoustic and linguistic models are optimized using machine learning methods like supervised learning or sequence-to-sequence models. To reduce recognition mistakes, this entails training on labeled data and modifying model parameters.ASR systems frequently go through rounds of iterative improvement. ASR performance may be improved over time by consistently updating language resources, incorporating user modifications, and improving model refinement with the base help of Recurrent Neural Networks (RNN).
The security department's monitoring system is trained with every approach, algorithm, technique and concept aforementionedin a sequential manner which ensures that the machine is well developed, skilled and incorporated with Artificial Intelligence.
The main reason for proposing the notion of facial recognition and motion detection in CCTV is to improve the monitoring system in educational facilities where the problem of a few pupils loitering about instead of attending class frequently arises. And the main issue is safeguarding against intruders, thefts, etc., not only in organizations but also in any confined space like a house, business, etc.
advantages of proposed invention,

Using this sophisticated autonomous surveillance systems that also include facial recognition and motion detection to replace human staff members helps to enhance productivity. Autonomous systems can work continuously, unlike human staff, who need breaks and rest periods.
To prevent loss or damage to the property, it may be necessary to report certain actions or strange activities right away, such as robberies or trespassing. In these circumstances, the suggested system operates effectively and provides the security department with prompt responses.
In addition, the surveillance system is incorporated with automatic speech recognition technique which the machine itself recognizes the audio and converts into its readable format and alerts the security division if any suspicious conversation is noticed. This surveillance system employs a number of different camera configurations, video management software, neural networks, data dissemination, database management system, and other technologies. This feature produces improved performance and a thorough footage of the confined space.

More effective than the conventional way of recognizing the person because it is automatic and takes much less time as the convolutional algorithms are finely used which enhances the video and image detection and extraction over a cluster of layers.
4 Claims and 2 Figures , Claims:The scope of the invention is defined by the following claims:
Claims:
1. The proposed invention integrated AI-based surveillance system (IASS ) comprises,
a) The model is mainly composed of two specifications namely facial recognition and motion detection for keen monitoring and enhanced surveillance.
b) In addition, to categorize persons as known and unknown, it employs object detection and recognition with the help of accessibility to the database.
c) Face recognition systems that employ AI analysis to accurately identify and validate persons based on their visual characteristics use convolutional neural networks (CNNs).
d) Using machine learning methods and computer vision techniques, AI analysis of motion involves finding and analysing motion patterns in video streams. By utilizing AI algorithms like CNNs and recurrent neural networks(RNNs), motion detection systems can precisely recognize and track moving objects.
2. As per claim 1, machine is well-trained using conventional algorithms and sophisticated methodologies, resulting in production of accurate and effective results, higher success rate, and reduced mistakes.
3. As per claim 1, Automatic speech recognition (ASR) is embedded into the machine in linkup with Recurrent Neural Networks so as to enable the audio output and making the surveillance system more effective and secure.
4. As per claim 1, in contrast to conventional methods, 3D CNN’s are incorporated which analyses multiple frames at a time resulting in increased efficiency, compatibility in real-time.