Preface

Current state-of-the-art technologies have opened up new opportunities in research in the areas of object detection and recognition of digital images and videos, robotics, neural networks, machine learning, stereo vision matching algorithms, soft computing, customer prediction, social media analysis, recommendation systems, and stereo vision. Therefore, this book has been designed to provide directions for those interested in researching and developing intelligent applications to detect an object and estimate depth. In addition to focusing on the performance of the system using high-performance computing techniques, a technical overview of certain tools, languages, libraries, frameworks, and APIs for developing applications is also given. More specifically, detection using stereo vision images/video from its developmental stage up till today, its possible applications, and general research problems relating to it are covered. Also presented are techniques and algorithms that satisfy the peculiar needs of stereo vision images along with emerging research opportunities through analysis of modern techniques being applied to intelligent systems.

Since both theoretical and practical aspects of the developments in this field of research are explored, including recent state-of-the-art technologies and research opportunities in the area of object detection, this book will act as a good reference for practitioners, students, and researchers. Briefly stated, since it is a pioneer reference in this particular field, it will be a significant source of information for researchers who have been longing for an integrated reference. It is ideally designed for researchers, academics, and post-graduate students seeking current research on emerging soft computing areas; and it can also be used by various universities as a textbook for graduate/post-graduate courses. Many professional societies, IT professionals, or organizations working in the field of robotics will also benefit from this book.

A chapter-by-chapter synopsis of the topics covered in this book follows:

• In Chapter 1, Deepti Nikumbh et al. present data conditioning techniques for medical imaging. Digital images have a tremendous influence on today's world and have become an essential component in the clinical medical field. Significant advancements in the processing of medical images, and improvements in diagnosis and analysis, have transformed medical imaging into one of today's hottest emerging fields for implementation and research. Image pre-processing techniques along with image segmentation and image processing algorithms are useful tools that pave the way for advancement in the medical field with wide applications such as cancer detection, fingerprint identification and many others using pattern matching, feature extraction and edge detection algorithms.
• In Chapter 2, Shravani Nimbolkar et al. present an analytical study for pneumonia detection using machine learning and deep learning techniques. This chapter studies different types of lung diseases and how their diagnosis can be aided using these techniques. Experimentation with different machine learning models, like CNN and MLP, and pre-trained architectures, like VGG16 and ResNet, are used to predict pneumonia from chest X-rays.
• In Chapter 3, Kavita R. Singh et al. explore the advanced application of a contamination monitoring system using IoT and GIS. Contamination/pollution is one of the biggest challenges where environmental issues are concerned. The authors analyze particular areas that are more contaminated/polluted in Nagpur City, Maharashtra, India, by calibrating the air quality index as an IoT-based air pollution monitoring framework and plotting the data using a geographical information system. Additionally, the data analysis, which is done with the help of Tableau software and different parameters like air quality index, temperature, etc., is provided to the end user through the android application.
• In Chapter 4, Rajani P.K. et al. present the new area of video error concealment using particle swarm optimization. Video transmission over wired or wireless channels, such as the internet, is the fastest growing area of research. The proposed method is a novel method in the spatio-temporal domain that can significantly improve the subjective and objective video quality. There are many algorithms for video error concealment. These optimized algorithms should be used for obtaining better video quality. Particle swarm optimization (PSO), which is one of the best optimized bio-in-spired algorithms, is used to conceal the errors in different video formats. Correlation is used for detection of errors in the videos and each error frame is concealed using PSO algorithm in MATLAB.
• In Chapter 5, Nalini Jagtap explores enhanced image fusion with guided filters. She proposes the modified guided filtering approach called "novel guided filtering" to overcome blurring and ringing effects. The primary step in this approach is to design the guidance image and generate the base and complex components based on that image. The edge detection operator plays a significant role in deciding the guidance image. The focus map is generated using low-rank representation, which is based on a detailed part of the original image. The built-in characteristic of removing ringing and blurring effects using LRR helps to develop artifact-free/noiseless detail-enhanced image fusion. First, guided filters are applied on a focus map; then the guided filter output is used to generate the resultant all-in-one fused image. In this case, ringing and blurring effects are removed using guided filters in the resultant fused image.
• In Chapter 6, Tejaswini Yesugade proposes deepfake detection using LSTM-based neural network. The rapid growth of social media and new developments in deep generative networks have improved the quality of creating more realistic fake videos, which are called deepfake videos. Such deepfake videos are used in politics to create political turmoil, for blackmail, and terrorism. To reduce the harm that can be done using such methods and prevent the spread of such fake images or videos, the author proposes a method that can detect such deepfakes and a new method to detect AI-generated fake videos using an algorithm such as CNN and LSTM. This method will detect deepfakes by using ResNext50 and LSTM algorithms, which have an accuracy of around 88%.
• In Chapter 7, Kavita Shinde et al. present various approaches for classification of fetal brain abnormalities with MRI images. Magnetic resonance imaging of fetuses allows doctors to observe brain abnormalities early on. Therefore, since nearly three out of every 1,000 fetuses have a brain anomaly, it is necessary to determine and categorize them at an earlier stage. The literature survey finds less work is involved in the classification of abnormal fetal brain based on conventional methods of machine learning, while more related work is conducted for the segmentation and feature extraction using different techniques. In this chapter, the authors review different machine learning techniques used for the complete MRI processing chain, starting with image acquisition to its classification.
• In Chapter 8, Chinnaiah Kotadi et al. explore a method to analyze COVID-19 data using a machine learning algorithm. The authors analyze past COVID-19 data to raise awareness of COVID-19 second wave conditions and precautions against the delta variant. They also provide COVID-19 cases such as confirmed cases, cured patients' cases, and death rates in India. Also, by using a machine learning algorithm, the states of India in which the most cases and deaths occurred are provided.
• In Chapter 9, Manish Sharma et al. explore an intelligent recommendation system for evaluating teaching faculty performance using adaptive collaborative filtering. This system uses the deep learning model for the evaluation and enhancement of the performance of teachers in educational institutions. To give a recommendation framework, this work incorporates numerous elements such as student assessment, intake quality, innovative practices, experiential learning approaches, and so on. The dataset derived from an educational institute's ERP was used to train and test the proposed recommender. The performance of the proposed recommender system was evaluated using the real-time data of teachers and other stakeholders from an educational institute apart from some secondary parameters. The comparative analysis of various techniques along with the performance comparison based on accuracy, precision, and recall are well furnished.
• In Chapter 10, Manisha Blende et al. propose a virtual moratorium system. By using the proposed system, the banker will get all the information regarding a customer who has opted for the moratorium. The user will interact with the chatbot and submit the moratorium request and then chatbot will ask questions based on customers' responses. Rasa natural language processing (NLP) and Rasa natural language understanding algorithm will classify the intents from the user responses. Intents will be compared with predefined patterns to extract the specific data. These responses will be stored in the NoSQL (MongoDB) database, and these data will be shared with the banker, who will further analyze them. The main purpose of this study is to help the banker know whether the customer who has applied for the moratorium is genuine. With this system, both the customer and the banker will be able to save time and effort. The proposed system will allow...