The project focused at developing tools for an autonomous ground vehicle (robot) for weed detection in sugar beet fields using 3D image data. So far, predominantly 2D images have been used in research in weed detection. A prerequisite for use of 2D data is that the individual plants are well separated not occluding each other. When plants grow in size, overlapping canopies make this approach difficult to resolve plants and analyse their structures. In such a case, 3D data will be beneficial. In this way, this study will help in localized spray of chemicals, hence reducing the amount of chemicals used in farming using image processing and computer vision.
This PhD study was financed by Danish Council of Strategic Research under project ASETA (Adaptive Surveying and Early treatment of crops with a Team of Autonomous vehicles) which also involves aerial vehicles and aerial imaging.
Esben Skouboe Poulsen
Together with Professor Ole B. Jensen.
Title: Responsive Public Lighting
The goal of the project was to investigate how architects can employ sensor technologies and embedded computational processes to extract valuable information about temporal social and cultural matters, such as; gatherings, groupings, occupancy, attention and activity. The thesis addresses architectural response scenarios for a site-sensitive performative architecture. Through a series of experiments the project develops a hierarchical approach to performance based design. In specific this thesis will focus on adaptive processes which facilitate social exchange and aid integration among different social groups, for a better and reinforced place.
Part of the study is affiliated to the following project under Dansk Lys Innovation network
Anne Marie Skriver Hansen
Together with Professor Pirkko Raudaskovski
Title: Improvisation and co-expression in explorative digital music systems
During the last two decades a new genre of “casual games,” defined by Juul 2010, has become popular among players of all age groups. Players who are physically present play these games in a social setting. The game play is simple and provides fairly short and intense play. This dissertation has investigated a new genre of casual games that allows players to both collaborate and be creative in their expression. A series of music-oriented games and play environments was designed, implemented and investigated according to how two players, who are not musicians, coordinate musical expression amongst each other when they are given a number of creative restrictions in the sonic/musical material that they interact with. The results from each study served to formulate new forms of level design in open-ended music oriented games, as well as design opportunities and “guidelines” for the future development of new digital and interlinked musical instruments for novices. This especially concerns which pattern recognition algorithms that should be applied in different types of music systems that challenge and mediate collaboration in music improvisation.
Søren Tranbjerg Hansen, together with Professor Thomas Bak
Title: Robot Games for Elderly – A Case-Based Approach
The focus of the thesis is on the development of games for an autonomous, mobile robot based on algorithms using spatio-temporal information about player behaviour. Three types of games each using a different control strategy for use within elder care, are investigated. The first game is based on basic robot control which allow the robot to detect and follow a person. The second game consists in an adaptive game algorithm which gradually adjusts the game challenge to the mobility skills of the player based on spatio-temporal analysis. The last game allows multiple users to compete against the robot based on an AI system originally created for real-time strategic computer games.
Together with Professor Thomas Bak
Title: Mobile Robots in Human Environments : towards safe, comfortable and natural navigation
Traditionally robots have been assistant machines in factories and an ubiquitous part of science fiction movies. But within the last decade the robots have started to emerge in everyday human environments. Today they are in our everyday environment in the shape of for example vacuum cleaners, lawn mowers, toy pets or as assisting technologies for care giving. If we want robots to be an even larger and more integrated part of our everyday environments, they need to become more intelligent, and behave safe and natural to the humans in the environment. This thesis deals with making intelligent robotic devices capable of being a more natural and sociable actor in a human environment. More specific the emphasis is on safe and natural motion and navigation issues.
Title: Image Analysis for Above-Canopy Measurement of Leaf Area Index
This thesis contains studies that were based on the idea of developing a camera sensor for measuring structure parameters, especially leaf area index (LAI), in crop canopies. Such measurements might be used for graduated application of e.g. fungicides or nitrogen fertilizers. There are handheld optical LAI-sensors commercially available, but these sensors must be placed at ground level, making them poorly suited for continuous automated measurements. Therefore, this thesis considers the use of downlooking camera sensor placed above the crops.
The contributions of this thesis fall into three areas:
1. Development of image analysis methods for measuring gap fraction and leaf area index using ordinary colour cameras, where most previous studies have used an infrared image band. The methods are validated with field tests in cereal crops under different illumination conditions, where most previous studies have been constrained to diffuse illumination conditions.
2. Proposal of a combined mixture analysis model for hyperspectral images for the separation of first-order and higher-order reflection components, and development of a stochastic prediction model in order to make it possible to use such components for estimating canopy structure parameters.
3. Review of existing weighting schemes for combination of multiple exposures in high-dynamic range imaging (e.g. outdoor), a comparison of their noise properties, and the proposal of an optimal weighting scheme for sensors with a linear response function.
Title: Automatic Plant Annotation Using 3D Computer Vision
In this thesis 3D reconstruction was investigated for application in precision agriculture where previous work focused on low resolution index maps where each pixel represents an area in the field and the index represents an overall crop status in that area. 3D reconstructions of plants would allow for more detailed descriptions of the state of the crops analogous to the way humans evaluate crop health, i.e. by looking at the canopy structure and check for discolorations at specific locations on the plants.