摘要 :
With the latest advances in image sensor technology, cameras are able to generate video with tens of megapixels per frame. These high resolution videos streams offer great potential to be used in the surveillance domain. For groun...
展开
With the latest advances in image sensor technology, cameras are able to generate video with tens of megapixels per frame. These high resolution videos streams offer great potential to be used in the surveillance domain. For ground based systems, gigapixel streams are already used with great effect as illustrated by the ICME 2019 crowd counting challenge. However, for Unmanned Aerial Vehicles (UAVs), this vast stream of data exceeds the limit of transmission bandwidth to send this data back to the ground. On board data analysis and selection is thus required to use and benefit from high resolution cameras. This paper presents a result of the CAVIAR project, where a combination of hardware and algorithms was designed to answer the question: "how to exploit a high resolution high frame rate camera on board a UAV?'. With the associated size, weight and power limitations, we implement data reduction by deploying deep learning on hardware to find the relevant information and transmit it to an operator station. The proposed solution aims at employing the high resolution potential of the sensor only onto objects of interest. We encode and transmit the identified regions containing those objects of interest (ROI) at the original resolution and framerate, while also transmitting the downscaled background to provide context for an operator. We demonstrate using a 35 fps, 65 Megapixel camera that this set-up indeed saves considerable bandwidth while retaining all important video data at high quality at the same time.
收起
摘要 :
With the latest advances in image sensor technology, cameras are able to generate video with tens of megapixels per frame. These high resolution videos streams offer great potential to be used in the surveillance domain. For groun...
展开
With the latest advances in image sensor technology, cameras are able to generate video with tens of megapixels per frame. These high resolution videos streams offer great potential to be used in the surveillance domain. For ground based systems, gigapixel streams are already used with great effect as illustrated by the ICME 2019 crowd counting challenge. However, for Unmanned Aerial Vehicles (UAVs), this vast stream of data exceeds the limit of transmission bandwidth to send this data back to the ground. On board data analysis and selection is thus required to use and benefit from high resolution cameras. This paper presents a result of the CAVIAR project, where a combination of hardware and algorithms was designed to answer the question: "how to exploit a high resolution high frame rate camera on board a UAV?'. With the associated size, weight and power limitations, we implement data reduction by deploying deep learning on hardware to find the relevant information and transmit it to an operator station. The proposed solution aims at employing the high resolution potential of the sensor only onto objects of interest. We encode and transmit the identified regions containing those objects of interest (ROI) at the original resolution and framerate, while also transmitting the downscaled background to provide context for an operator. We demonstrate using a 35 fps, 65 Megapixel camera that this set-up indeed saves considerable bandwidth while retaining all important video data at high quality at the same time.
收起
摘要 :
In recent years advances in machine learning methods such as deep learning has led to signi cant improvements in
our ability to track people and vehicles, and to recognise speci c individuals. Such technology has enormous po-
tent...
展开
In recent years advances in machine learning methods such as deep learning has led to signi cant improvements in
our ability to track people and vehicles, and to recognise speci c individuals. Such technology has enormous po-
tential to enhance the performance of image-based security systems. However, wide-spread use of such technology
has important legal and ethical implications, not least for individuals right to privacy. In this paper, we describe
a technological approach to balance the two competing goals of system e cacy and privacy. We describe a
methodology for constructing a \goal-function" that re
ects the operators preferences for detection performance
and anonymity. This goal function is combined with an image-processing system that provides tracking and
threat assessment functionality and a decision-making framework that assesses the potential value gained by
providing the operator with de-anonymized images. The framework provides a probabilistic approach combining
user preferences, world state model, possible user actions and threat mitigation e ectiveness, and suggests the
user action with the largest estimated utility. We show results of operating the system in a perimeter-protection
scenario.
收起
摘要 :
In recent years advances in machine learning methods such as deep learning has led to signicant improvements in
our ability to track people and vehicles, and to recognise specic individuals. Such technology has enormous po-
te...
展开
In recent years advances in machine learning methods such as deep learning has led to signicant improvements in
our ability to track people and vehicles, and to recognise specic individuals. Such technology has enormous po-
tential to enhance the performance of image-based security systems. However, wide-spread use of such technology
has important legal and ethical implications, not least for individuals right to privacy. In this paper, we describe
a technological approach to balance the two competing goals of system ecacy and privacy. We describe a
methodology for constructing a \goal-function" that re
ects the operators preferences for detection performance
and anonymity. This goal function is combined with an image-processing system that provides tracking and
threat assessment functionality and a decision-making framework that assesses the potential value gained by
providing the operator with de-anonymized images. The framework provides a probabilistic approach combining
user preferences, world state model, possible user actions and threat mitigation eectiveness, and suggests the
user action with the largest estimated utility. We show results of operating the system in a perimeter-protection
scenario.
收起
摘要 :
Elektrotraktion bietet im Rahmen der aktuellen Klimadiskussion eine okologisch sinnvolle Alternative zu verbrennungsmotorisch betriebenen Fahrzeugen. Unter dieser Technologie werden in der Kraftstoff- und Antriebsstrategie von Vol...
展开
Elektrotraktion bietet im Rahmen der aktuellen Klimadiskussion eine okologisch sinnvolle Alternative zu verbrennungsmotorisch betriebenen Fahrzeugen. Unter dieser Technologie werden in der Kraftstoff- und Antriebsstrategie von Volkswagen sowohl der rein batterieelektrische Antrieb als auch der Brennstoffzellenantrieb zusammengefasst, es handelt sich bei beiden um elektromotorische Antriebe, die sich lediglich durch die Art der Energiebereitstellung unterscheiden. Die in automobilen Anwendungen vorzugsweise eingesetzten Niedertemperatur-PEM-Brennstoffzellen benotigen fur die elektrochemische Energiewandlung die Elemente Wasserstoff und Sauerstoff. Letzterer wird dabei aus der Umgebungsluft mit Hilfe eines Aufladeaggregats der Brennstoffzelle zugefuhrt. In den vergangenen Fahrzeuggenerationen HyMotion1 bis HyMotion3 der Volkswagen AG wurden bisher Schraubenkompressoren zur Bereitstellung des Luftmassenstroms bei den geforderten Druckverhaltnissen eingesetzt [1]. Die Betriebserfahrungen mit den Schraubenkompressoren in der HyMotion3-Fahrzeugflotte zeigen als Vorteile ein zuverlassiges Betriebsverhalten und die Abdeckung eines sehr breiten Kennfeldbereiches auf. Demgegenuber stehen die ungunstige Akustik, die relativ hohen Fertigungskosten und die vergleichsweise geringe Dynamik bei wechselnden Lastanforderungen, bedingt durch die Tragheitsmomente der Schraubenrotoren. Unter diesen Gesichtspunkten ist in der nachsten Fahrzeuggeneration zur Systemoptimierung der Einsatz von Turbomaschinen vorgesehen. Der Artikel stellt im folgenden die Aufladetechnologien gegenuber und beschreibt Betriebserfahrungen, Vor- und Nachteile und die sich daraus ergebenden Aufgabenstellungen, die bearbeitet werden mussen, damit Schraubenmaschinen eine Renaissance als Aufladeaggregate in Brennstoffzellenfahrzeugen erleben.
收起
摘要 :
At Amsterdam Airport Schiphol, innovative technology is developed to support automated predictive behavioural profiling of deviant behaviour from CCTV camera footage. The Royal Netherlands Marechaussee expects predictive behaviour...
展开
At Amsterdam Airport Schiphol, innovative technology is developed to support automated predictive behavioural profiling of deviant behaviour from CCTV camera footage. The Royal Netherlands Marechaussee expects predictive behavioural profiling to become a valuable asset to performing its tasks in the near future.
收起
摘要 :
An excimer laser source was used for shock treatment of metals. The treated specimens were materials applied in automotive and aeronautical industry. The process was investigated with physical methods and optimized in terms of hig...
展开
An excimer laser source was used for shock treatment of metals. The treated specimens were materials applied in automotive and aeronautical industry. The process was investigated with physical methods and optimized in terms of high surface pressures. The results of the shock treatment concerns the modifications of the mechanical properties such as roughness, hardness and residual stresses.
收起
摘要 :
The contribution of this paper is a search engine that recognizes and describes 48 human actions in realistic videos. The core algorithms have been published recently, from the early visual processing (Bouma, 2012), discriminative...
展开
The contribution of this paper is a search engine that recognizes and describes 48 human actions in realistic videos. The core algorithms have been published recently, from the early visual processing (Bouma, 2012), discriminative recognition (Burghouts, 2012) and textual description (Hanckmann, 2012) of 48 human actions. We summarize the key algorithms and specify their performance. The novelty of this paper is that we integrate these algorithms into a search engine. In this paper, we add an algorithm that finds the relevant spatio-temporal regions in the video, which is the input for the early visual processing. As a result, metadata is produced by the recognition and description algorithms. The meta-data is filtered by a novel algorithm that selects only the most informative parts of the video. We demonstrate the power of our search engine by retrieving relevant parts of the video based on three different queries. The search results indicate where specific events occurred, and which actors and objects were involved. We show that events can be successfully retrieved and inspected by usage of the proposed search engine.
收起
摘要 :
Shadow detection is an important aspect of foreground/background classification. Many techniques exist, most of them assuming that casting a shadow results only in a change in intensity and no change in color. In this paper we sho...
展开
Shadow detection is an important aspect of foreground/background classification. Many techniques exist, most of them assuming that casting a shadow results only in a change in intensity and no change in color. In this paper we show that in most practical indoor and outdoor situations there will also be a color shift. We propose an algorithm for estimating this color shift from the images, and using it to remove shadow pixels. The proposed algorithm is compared experimentally to an existing algorithm using real image sequences. Results show a significant improvement of performance.
收起
摘要 :
Infrared (IR) cameras are often used in a vehicle-based multi-sensor platform for landmine detection. Additional to thermal contrasts, an IR polarimetric sensor also measures surface properties and therefore has the potential of i...
展开
Infrared (IR) cameras are often used in a vehicle-based multi-sensor platform for landmine detection. Additional to thermal contrasts, an IR polarimetric sensor also measures surface properties and therefore has the potential of increased detection performance. We have developed a polarimetric IR setup, which has to be used in a forward-looking manner. This paper describes all the steps to reach detection performance. The first step is the acquisition of the polarimetric IR image data. The next step is the pre-processing to (re)construct polarimetric images. A subsequent segmentation step is made to identify objects. Features, like intensity, reflectivity and shape, of these objects are measured. For independent performance analyses, the data set is divided into a training and evaluation section. A classifier is trained on the training section and evaluated on the classification section. The detection performance of the forward-looking IR camera is shown in receiver operator characteristics (ROC) curves.
收起