In diesem Beitrag wird die Problematik der effizienten Zuordnung von Sensorressourcen zum Zweck der Objektklassifikation in Straßenverkehrsszenen behandelt. Vorgestellt wird ein System bestehend aus einer niedrig auflösenden 3-D Kamera und zwei hochauflösenden optischen Sensoren, das durch eine kontextsensible Nutzenoptimierung gesteuert wird.
In this paper, the problem of efficiently allocating sensor-resources in order to collect to classify objects in dynamic road traffic scenes is addressed. A system using a low-resolution 3-D camera, and two high-resolution optical sensors controlled by a context aware utility optimisation instance is proposed. The use of contextual knowledge to improve cue combination for gaze direction selection is proposed. Using intensity, range and motion as features in order to perform object / background recognition, saliency detection, and occlusion detection a syntactical set of cues is combined using a utility optimisation scheme. Contextual awareness is gained by determining the present road environment using a map and augmented with current classification results. The calculation of cues and the gaze direction selection process are shown to perform in real-time using a 2 GHz Intel Pentium CPU. Further work will cover the quantitative evaluation of gaze direction selection using scenes with labelled ground truth as well as the quantitative evaluation of cue combination improvement using contextual knowledge.
Proactive sensor-resource allocation using optical sensors
Proaktive Sensorsteuerung optischer Sensoren
2008
9 Seiten, 4 Bilder, 1 Tabelle, 12 Quellen
Aufsatz (Konferenz)
Englisch
Proactive Sensor-Resource Allocating using Optical Sensors
British Library Conference Proceedings | 2008
|Europäisches Patentamt | 2023
|Europäisches Patentamt | 2022
|Using a Proactive Sensor-System in the Distributed Environment Model
British Library Conference Proceedings | 2008
|