Publicação
Optimized video retrieval for interior vehicle monitoring
| Resumo: | With the rapid growth in the amount of video data, an increasing need for efficient video retrieval systems has become an important problem in the multimedia management topic. Despite having a long past, the increase in file size of video collections, caused mostly by the increase of video resolution and quantity of videos, originated a big push for applying Machine Learning on the video retrieval subject. In today’s world, when dealing with Big Data, it’s unfeasible to still rely on video metadata and manually annotated videos to provide an accurate video retrieval engine, seeing as the sheer quantity of videos overwhelms an inept search and browse system, unable to provide the video the user wants. Therefore, by relying on machine algorithms to accurately mass tag the video collection we achieve great improvements. The process of allocating the video information to the video retrieval framework is severely less time consuming and the viewer has at his disposal more precise and semantically accurate filters. This in turn, drastically reduces the quantity of redundant videos that are pulled from the user’s queries. Another way to also ease the time it takes to analyze an immense quantity of videos, is by summarizing the content that is present on them. Condensing dozens of hours, pulled from one or more video streams, into a more accessible source of information that displays the most relevant data, is considerably a more efficient viewing experience for the user as it unburdens him of the task of surveying a grotesque amount of media content. The main focus of this thesis is to implement a video summarization method for recapping footage from the interior of a vehicle, that will be integrated on a video retrieval platform that is also being developed in parallel. |
|---|---|
| Autores principais: | Dias, Diogo Barroso |
| Assunto: | Video retrieval Video summarization Histogram Disparity minimization Greedy algorithm Machine learning Recuperação de vídeo Sumarização de vídeo Histograma Minimização de disparidades Algorítmo ganancioso Aprendizagem automática |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | With the rapid growth in the amount of video data, an increasing need for efficient video retrieval systems has become an important problem in the multimedia management topic. Despite having a long past, the increase in file size of video collections, caused mostly by the increase of video resolution and quantity of videos, originated a big push for applying Machine Learning on the video retrieval subject. In today’s world, when dealing with Big Data, it’s unfeasible to still rely on video metadata and manually annotated videos to provide an accurate video retrieval engine, seeing as the sheer quantity of videos overwhelms an inept search and browse system, unable to provide the video the user wants. Therefore, by relying on machine algorithms to accurately mass tag the video collection we achieve great improvements. The process of allocating the video information to the video retrieval framework is severely less time consuming and the viewer has at his disposal more precise and semantically accurate filters. This in turn, drastically reduces the quantity of redundant videos that are pulled from the user’s queries. Another way to also ease the time it takes to analyze an immense quantity of videos, is by summarizing the content that is present on them. Condensing dozens of hours, pulled from one or more video streams, into a more accessible source of information that displays the most relevant data, is considerably a more efficient viewing experience for the user as it unburdens him of the task of surveying a grotesque amount of media content. The main focus of this thesis is to implement a video summarization method for recapping footage from the interior of a vehicle, that will be integrated on a video retrieval platform that is also being developed in parallel. |
|---|