TY  - THES
AU  - Ivan Huerta
ED  - Jordi Gonzalez
ED  - Xavier Roca
PY  - 2010//
TI  - Foreground Object Segmentation and Shadow Detection for Video Sequences in Uncontrolled Environments
PB  - Ediciones Graficas Rey
N2  - This Thesis is mainly divided in two parts. The first one presents a study of motionsegmentation problems. Based on this study, a novel algorithm for mobile-objectsegmentation from a static background scene is also presented. This approach isdemonstrated robust and accurate under most of the common problems in motionsegmentation. The second one tackles the problem of shadows in depth. Firstly, abottom-up approach based on a chromatic shadow detector is presented to deal withumbra shadows. Secondly, a top-down approach based on a tracking system has beendeveloped in order to enhance the chromatic shadow detection.In our first contribution, a case analysis of motion segmentation problems is presented by taking into account the problems associated with different cues, namelycolour, edge and intensity. Our second contribution is a hybrid architecture whichhandles the main problems observed in such a case analysis, by fusing (i) the knowledge from these three cues and (ii) a temporal difference algorithm. On the one hand,we enhance the colour and edge models to solve both global/local illumination changes(shadows and highlights) and camouflage in intensity. In addition, local information isexploited to cope with a very challenging problem such as the camouflage in chroma.On the other hand, the intensity cue is also applied when colour and edge cues are notavailable, such as when beyond the dynamic range. Additionally, temporal differenceis included to segment motion when these three cues are not available, such as thatbackground not visible during the training period. Lastly, the approach is enhancedfor allowing ghost detection. As a result, our approach obtains very accurate and robust motion segmentation in both indoor and outdoor scenarios, as quantitatively andqualitatively demonstrated in the experimental results, by comparing our approachwith most best-known state-of-the-art approaches.Motion Segmentation has to deal with shadows to avoid distortions when detectingmoving objects. Most segmentation approaches dealing with shadow detection aretypically restricted to penumbra shadows. Therefore, such techniques cannot copewell with umbra shadows. Consequently, umbra shadows are usually detected as partof moving objects.Firstly, a bottom-up approach for detection and removal of chromatic movingshadows in surveillance scenarios is proposed. Secondly, a top-down approach basedon kalman filters to detect and track shadows has been developed in order to enhancethe chromatic shadow detection. In the Bottom-up part, the shadow detection approach applies a novel technique based on gradient and colour models for separatingchromatic moving shadows from moving objects.Well-known colour and gradient models are extended and improved into an invariant colour cone model and an invariant gradient model, respectively, to performautomatic segmentation while detecting potential shadows. Hereafter, the regions corresponding to potential shadows are grouped by considering ”a bluish effect” and anedge partitioning. Lastly, (i) temporal similarities between local gradient structuresand (ii) spatial similarities between chrominance angle and brightness distortions areanalysed for all potential shadow regions in order to finally identify umbra shadows.In the top-down process, after detection of objects and shadows both are trackedusing Kalman filters, in order to enhance the chromatic shadow detection, when itfails to detect a shadow. Firstly, this implies a data association between the blobs(foreground and shadow) and Kalman filters. Secondly, an event analysis of the different data association cases is performed, and occlusion handling is managed by aProbabilistic Appearance Model (PAM). Based on this association, temporal consistency is looked for the association between foregrounds and shadows and theirrespective Kalman Filters. From this association several cases are studied, as a resultlost chromatic shadows are correctly detected. Finally, the tracking results are usedas feedback to improve the shadow and object detection.Unlike other approaches, our method does not make any a-priori assumptionsabout camera location, surface geometries, surface textures, shapes and types ofshadows, objects, and background. Experimental results show the performance andaccuracy of our approach in different shadowed materials and illumination conditions.
SN  - 978-84-937261-3-3
N1  - exported from refbase (http://refbase.cvc.uab.es/show.php?record=1332), last updated on Fri, 17 Dec 2021 13:59:00 +0100
ID  - Ivan Huerta2010
U1  - Ph.D. thesis
ER  -