Application of Fuzzy Approximation Method in Pattern Recognition Using Deep Learning Neural Networks and Artificial Intelligence for Surveillance
M. Geethalakshmi, V. Sriram, Vakkalagadda Drishti Rao · 2024
This paper deals with the idea of image processing techniques in surveillance or security systems. The main purpose of this study is to build a constant monitoring system with the help of a deep learning neural network and artificial intelligence by introducing a new fuzzification technique called Fuzzy Approximation Method (FAM). Even though there are many difficulties in image processing, the challenging task is to fix the variations of backgrounds during monitoring. Though different contributions on this issue have been made, the surveillance system still faces a deficit of efficiency. This study helps to give more accurate clarity of the image by processing or converting the data and values of the plots in the form of an image. The processed image is scanned in various stages of hidden layers and brings it to a form of plotting points, and the graph is generated by using MATLAB tool through artificial intelligence. Step-by-step algorithm is followed throughout the entire process. The processed values are again stored in a variable as a resultant value by using a neural network in addition to the new technique. The same process of plotting is repeated once in a while a new image is processed or scanned in order to get a clear picture. The value obtained in this fuzzy process along with the neural network will be compared along with the results of original or actual value in each step so that it will be accepted when it is equal and denied if it is not matching. For tracing the identity and to maintain the originality by reducing duplication, a secondary analysis is done in order to identify the correct image of the user. This will ensure the accuracy of the anomaly user when using the image key, it will send the result as the data found to be invalid. The combined effect of artificial intelligence with neural networks through fuzzy approximation methods helps to improve the process to work more quickly and efficiently. By implementing this idea in surveillance or security systems it leads to accuracy in the intensity while capturing the image. The execution of FAM is necessary to make the surveillance or the security system work faster and more efficiently, and it reduces the computation time. Even this work can be extended with the support of video and audio facility for dividing the recordings into frames of images and the same image will be related and the overall resulting value is stored to improve the clarity. Finally, background variation is rectified through this image processing. Thus, this proposed study is definitely a better alternative for pattern classification.