People counting is a challenging task with many applications. We propose a method with a fixed stereo camera that is based on projecting a template onto the depth image. The method was tested on a challenging outdoor dataset with good results and runs in real time.
DOCUMENT
In this paper we propose a head detection method using range data from a stereo camera. The method is based on a technique that has been introduced in the domain of voxel data. For application in stereo cameras, the technique is extended (1) to be applicable to stereo data, and (2) to be robust with regard to noise and variation in environmental settings. The method consists of foreground selection, head detection, and blob separation, and, to improve results in case of misdetections, incorporates a means for people tracking. It is tested in experiments with actual stereo data, gathered from three distinct real-life scenarios. Experimental results show that the proposed method performs well in terms of both precision and recall. In addition, the method was shown to perform well in highly crowded situations. From our results, we may conclude that the proposed method provides a strong basis for head detection in applications that utilise stereo cameras.
MULTIFILE
This study presents an automated method for detecting and measuring the apex head thickness of tomato plants, a critical phenotypic trait associated with plant health, fruit development, and yield forecasting. Due to the apex's sensitivity to physical contact, non-invasive monitoring is essential. This paper addresses the demand for automated, contactless systems among Dutch growers. Our approach integrates deep learning models (YOLO and Faster RCNN) with RGB-D camera imaging to enable accurate, scalable, and non-invasive measurement in greenhouse environments. A dataset of 600 RGB-D images captured in a controlled greenhouse, was fully preprocessed, annotated, and augmented for optimal training. Experimental results show that YOLOv8n achieved superior performance with a precision of 91.2 %, recall of 86.7 %, and an Intersection over Union (IoU) score of 89.4 %. Other models, such as YOLOv9t, YOLOv10n, YOLOv11n, and Faster RCNN, demonstrated lower precision scores of 83.6 %, 74.6 %, 75.4 %, and 78 %, respectively. Their IoU scores were also lower, indicating less reliable detection. This research establishes a robust, real-time method for precision agriculture through automated apex head thickness measurement.
DOCUMENT
BackgroundA valuable opportunity for reducing the fall incidence in hospitals, is alerting nurses when a patient is about to fall. For such a fall prevention system, more knowledge is needed on what occurs right before a fall. This can be achieved with a stereo camera that automatically detects (and records) dangerous situations.MethodsInpatients with a high risk of falling are selected for inclusion. A fall-risk questionnaire is administered and falls are logged during their stay. A stereo-camera (3D BRAVO-EagleEye system) is mounted in the ceiling and monitors the bed with surroundings. A baseline recording is made to improve the algorithms behind the alert system. When a fall or dangerous situation is detected, monitoring data preceding the incident is stored. Data is analyzed to assess 1) the quality of the system and 2) the prevalence of dangerous situations. Interviews with senior nurses are included in the evaluation.ResultsData collection is ongoing (Currently n=18; falls=1), and currently consists of ±62 hour of baseline recordings and ±24 hour of event-based recordings. These recordings include false positives as well as actual high risk situations.ConclusionsDespite the initial enthusiasm of the participating departments, inclusion of participants is slow, and the number of falls lower than expected. Possible explanations for this have been discussed with the involved senior nurses. With the monitoring data we gained more insight into the occurrence of dangerous situations, but to be able to reliably predict falls, more data on actual fallsshould be recorded.
DOCUMENT
Het project PreciSIAlandbouw heeft precisielandbouwtechnieken ontwikkeld en gevalideerd op vijf thema's: sensortechnologie, kennis en advies, robotisering, digitalisering, en verdienmodellen. Dit rapport bevat de resultaten van robotisering. Er zijn modules ontwikkeld om gewas en onkruid te onderscheiden en locaties van plantdetails nauwkeurig te bepalen.Hogeschool Saxion, lectoraat Lectoraat Smart Mechatronics and Robotics
DOCUMENT
Autonomous driving in public roads requires precise localization within the range of few centimeters. Even the best current precise localization system based on the Global Navigation Satellite System (GNSS) can not always reach this level of precision, especially in an urban environment, where the signal is disturbed by surrounding buildings and artifacts. Laser range finder and stereo vision have been successfully used for obstacle detection, mapping and localization to solve the autonomous driving problem. Unfortunately, Light Detection and Ranging (LIDARs) are very expensive sensors and stereo vision requires powerful dedicated hardware to process the cameras information. In this context, this article presents a low-cost architecture of sensors and data fusion algorithm capable of autonomous driving in narrow two-way roads. Our approach exploits a combination of a short-range visual lane marking detector and a dead reckoning system to build a long and precise perception of the lane markings in the vehicle’s backwards. This information is used to localize the vehicle in a map, that also contains the reference trajectory for autonomous driving. Experimental results show the successful application of the proposed system on a real autonomous driving situation.
DOCUMENT
In this paper, we address the problem of people detection and tracking in crowded scenes using range cameras. We propose a new method for people detection and localisation based on the combination of background modelling and template matching. The method uses an adaptive background model in the range domain to characterise the scene without people. Then a 3D template is placed in possible people locations by projecting it in the background to reconstruct a range image that is most similar to the observed range image. We tested the method on a challenging outdoor dataset and compared it to two methods that each shares one characteristic with the proposed method: a similar template-based method that works in 2D and a well-known baseline method that works in the range domain. Our method performs significantly better, does not deteriorate in crowded environments and runs in real time.
DOCUMENT
Wat zijn belangrijke succesfactoren om onderzoek, onderwijs en ondernemen bij elkaar te brengen, zó dat 'het klikt'. De uitdaging voor de toekomst van bedrijven in de smart factoryligt bij data science: het omzetten van ruwe (sensor) data naar (zinnige) informatie en kennis, waarmee producten en diensten verbeterd kunnen worden. Tevens programma van het symposium t.g.l. inauguratie 3 december 2015
MULTIFILE
De auto is niet meer weg te denken in onze huidige westerse maatschappij en bezet een belangrijke plaats in zowel ons economische als sociale leven. Hoewel Nederland al een van de meest verkeersveilige landen ter wereld is, waren er toch nog 811 verkeersdoden in 2006. Als we ons echter realiseren dat dit slechts een kwart is van de ruim 3200 verkeersdoden in 1972, is sindsdien al veel bereikt. De Nederlandse overheid streeft naar een verdere reductie tot minder dat 580 verkeersdoden in 2020. De daarvoor noodzakelijke verbeterde verkeersveiligheid zal voor een groot deel moeten komen uit nieuwe voertuigtechnologie die ongevallen helpt voorkomen (actieve veiligheid) en de gevolgen ervan beperkt (passieve veiligheid). Een auto veilig door het hedendaagse verkeer loodsen is geen eenvoudige taak, zeker niet onder slechte weersomstandigheden en bij complexe of onoverzichtelijke verkeerssituaties. Het is dan ook niet verwonderlijk dat bij het overgrote deel van de verkeersongevallen de oorzaak, minstens ten dele, bij een menselijke fout ligt. Intelligente voertuigsystemen, die met behulp van aan het voertuig verbonden omgevingssensoren het verkeer rond het voertuig monitoren, kunnen de bestuurder assisteren. Als er zich geen bijzonderheden voordoen is de bestuurder het meest gebaat bij informatieve- en comfortverhogende systemen. Als er een gevaarlijke situatie dreigt te ontstaan, komen de veiligheidssystemen in beeld. Naarmate de kans op een ongeval toeneemt, lijkt een grotere mate van ondersteuning (van waarschuwen, via assisteren tot interveniëren) gewenst. Vanwege hun veiligheidskritische karakter moeten actieve veiligheidssystemen voldoen aan hoge eisen ten aanzien van prestatie (hoge nauwkeurigheid), robuustheid (weersomstandigheden en wegcondities) en betrouwbaarheid. Hier liggen enorme uitdagingen in zowel het ontwerp als de evaluatie van dergelijke systemen waaraan het lectoraat Automotive control van Fontys Hogescholen door praktijkgericht onderzoek en vraaggestuurd onderwijs wil bijdragen.
DOCUMENT
