Guest Editorial: Environmental Multimedia Retrieval

摘要: The rapid advancements of digital technologies, as well the progress and wide availability cameras sensors have resulted in a great increase multimedia data production worldwide. This is also case for that describe state environment, which include huge amounts streams from model systems, dedicated stations amateur sensors, visual environmental information, such heatmaps forest satellite images. In parallel, success citizen sciences social networking tools has fostered emergence large structured communities nature observers (e.g., e-bird, Tela Botanica, etc.), who started to produce outstanding collections biodiversity records. Citizens become increasingly aware important role weather forecast, air quality, life species distributions) play on health issues allergies), variety other human activities agriculture, trip planning). addition, are very understanding phenomena, greenhouse effect, global warming climate change. Therefore, there an increasing need development advanced techniques analysing, interpreting aggregating provided formats. will allow generation reliable measurements, personalised applications take into account environment personal conditions preferences [5]. Furthermore, it should be noted accurate timely knowledge living essential sustainable humanity conservation. Despite fact number analysis been developed specifically extracting events behaviours human-centered general purpose applications, sports, movies, surveillance, relatively little attention paid analysis, retrieval interpretation information content. Only recent projects PESCaDO1 Pl@ntNet2 PASODOBLE3 dealt with developing innovative services investigated extraction semantic encoded format weather, pollen forecasts or citizen’s this context, works deal heatmap forecast [3], plant identification [2], underwater [4] monitoring atmosphere [1]. In context thematic issue focuses processing, indexing, fusion meteorological multimedia, classification recognition plants biological applications. special consists 7 papers. The first paper deals “Focussed Crawling Environmental Web Resources Based Combination Multimedia Evidence” ( 10.1007/s11042-015-2624-3). work proposes classifier-guided focussed crawling approach estimates relevance hyperlink unvisited resource based combination textual evidence. proposed applied towards discovery resources provide quality measurements forecasts. Then, includes “Environmental using AirMerge system” 10.1007/s11042-015-2604-7). platform allows multiple heterogeneous chemical sources continuously collected unified repository. results demonstrate potential being areas, image-based needed. Another challenging task research identification. “Semantic-based automatic structuring leaf images identification” 10.1007/s11042-015-2603-8) proposed. First, process performed by defining geometric parameters botanical definitions. Then, texture contour descriptors followed KNN classifier. Experiments carried out Scan Pictures ImageCLEF 2011 dataset. In we present “Plant identification: Man vs. Machine On beyond LifeCLEF 2014 challenge” 10.1007/s11042-015-2607-4). reports large-scale experiment aimed at evaluating how state-of-art computer vision systems perform identifying compared expertise. A subset dataset used within challenge was employed. One main outcomes machines still far beating best expert botanists. However, machine runs competing experienced botanists do clearly outperform beginners inexperienced test subjects. shows performances automated promising might open door new ecological surveillance systems. The next “Point-based Medialness 2D Shape Description Identification” 10.1007/s11042-015-2605-6). work, perception-based medial point description natural form framework shape representation proposed, can then efficiently matching tasks. Also, robust algorithm designed finds most relevant targets database templates comparing feature points scale, rotation translation invariant way. performance method tested several databases. The applies water landscape. Specifically “Fine-Grained Object Recognition Underwater Visual Data” 10.1007/s11042-015-2601-x) determining fish low-quality image video data. label propagation presented, able transfer labels annotated set 20 million order achieve variability appearance. Finally, interesting application, seismic damage assessment. “An assessment 10.1007/s11042-015-2602-9) called IDEAS evaluate inside buildings. compares taken building before after earthquake maps Mercalli intensity scale. investigate effectiveness accuracy IDEAS, over forty pairs closed-circuit television (CCTV) Youtube website used. These seven papers challenges therefore they appealing both experts field, those wish snapshot current breadth research.