Unique framework for video segmentation, and categorization applicable to traffic and medical environments

自从数字媒体出现以来，它已在许多方面被用于自动化人工工作，并在不同的应用领域带来效率。在其他媒体中，视频处理因其表现力而一直是最受欢迎的技术之一。视频几乎在世界各地的广告、监控、交通管制和娱乐中使用。因此，视觉数据的自动处理在研究和技术方面也越来越受欢迎。该项目的主要目标是场景理解，这是自动视频处理的核心要素。 场景理解是指自动检测和识别视频中存在的对象，目的是用相应的搜索标签对视频进行注释。自动场景理解已经在交通监控、监控、基于上下文的视频检索、视频分类和医学视频分析等领域得到了应用。视频中有意义对象的分割是场景理解过程的前一步。 因此，我们提出了一种独特的用于场景理解的视频分割和分类框架。需要相当多的背景知识和专门知识，才能促进和改进这一领域的现有工作。通过多年的背景发展和研究，我们得出的结论是，需要通过将目标划分为不同的任务来制定框架，以实现一些短期目标，总体上代表构建场景理解框架的最终长期目标。我们估计将任命九(9)名专职HQP来单独完成每项任务，他们的努力最终实现了最终目标。短期目标是设计该框架，并针对不同环境进行验证和验证。长远目标是特别针对交通和医疗环境的应用领域，因为这两个领域对公众安全和健康有重大影响。这一新颖的框架也将为更广泛的场景理解奠定基础。

Since the advent of digital media, it has been utilized in numerous ways to automate manual work and bring about efficiencies in different application areas. Video processing, among other media, has been one of the most sought-after technologies due to its power of representation. Videos are used in advertising, surveillance, traffic regulation, and entertainment virtually everywhere in the world. Thus, automatic processing of visual data has also seen increased popularity in research and technology. The main target of this project is in the sector of scene understanding, which is a central element of automatic video processing. Scene understanding refers to the automatic detection and identification of the objects present in videos with the aim of annotating videos with corresponding search tags. Automatic scene understanding has been applied in several sectors, namely traffic monitoring, surveillance, context-based video retrieval, video categorization, and medical video analysis. The segmentation of meaningful objects in videos is the step prior to the process of scene understanding. Therefore, we are proposing a unique framework for video segmentation and categorization for scene understanding. Considerable background knowledge and expertise is required to contribute to and improve upon existing work in this area. Through years of background development and research, we have concluded that the framework needs to be developed by dividing the target into distinct tasks to achieve a number of short-term objectives, holistically representing the final long-term objective of constructing a framework for scene understanding. We estimated nine (9) dedicated HQPs to be appointed for the accomplishment of each task individually, with their efforts culminating in achieving the final goal. The short-term objective is to design the framework, verify and validate it for different environments. The long-term objective is to specifically target the application areas of traffic and medical environments since these two areas have major impact on public safety and health. This novel framework would also establish a stepping stone towards more general scene understanding.