Master of Science / Department of Computing and Information Sciences / David A. Gustafson / An interest point detection scheme is presented that is comparable in quality to existing methods, but can be performed much faster. The detection is based on a straightforward color analysis at a coarse granularity. A 3x3 grid of squares is centered on the candidate point, so that the candidate point corresponds to the middle square. If the color of the center region is inhomogeneous with all of the surrounding regions, the point is labeled as interesting. A point will also be labeled as interesting if a minority of the surrounding squares are homogeneous, and arranged in an appropriate pattern. Testing confirms that this detection scheme is much faster than the state-of-the-art. It is also repeatable, even under different viewing conditions. The detector is robust with respect to changes in viewpoint, lighting, zoom, and to a certain extent, rotation.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

This project is developed with the University of Skövde and Volvo GTO. Purpose of the project is to complement and facilitate the quality insurance when gluing the engine frame. Quality defects in today’s industry is a major concern due to how costly it is to fix them. With competition rising and quality demands increasing, companies are looking for new and more efficient ways to ensure quality. Collaborative robots is a rising and unexplored technology in most industries. It is an upcoming field with great flexibility that could solve many issues and can assist its processes that are difficult to automate. The project aims to investigate if it is possible and beneficial to implement a vision system on a collaborative robot which ensures quality. Also, investigate if the collaborative robot could work with other tasks as well. This project also includes training and learning an artificial network with CAD generated models and real-life prototypes. The project had a lot of challenges with both training the AI and how the robot would communicate with it. The final results stated that a collaborative robot more specific UR10e could work with machine vision. This solution was based on using a camera which was compatible with the built-in robot software. However, this does not mean that other type of cameras cannot be used for this type of functions as well. Using machine vision based on artificial intelligence is a valid solution but requires further development and training to get a software function working in industry. Working with collaborative robots could change the industry for the better in many ways. Implementing collaborative robots could ease the work for the operators to aid in heavy lifting and repetitive work. Being able to combine a collaborative robot with a vision system could increase productivity and economic benefits.

Correlation-based translated-feature finding techniques are fast and effective in identifying targets in test images despite unknown translation. Information involving both translation and in-plane orientation of targets, however, is important in many industrial machine vision applications such as manufacturing and quality assurance. A traditional correlation based technique that expands the search criteria to include in-plane orientation is based upon use of a bank of filters that each implement a feature finding operation for one rotation of the target. This computational complexity of this approach is inversely proportional to the resolution of the orientation estimate. This thesis develops a correlation based method for translation and in-plane orientation feature finding that requires only two underlying correlation filter operations. A composite filter is constructed from a specially arranged and complex weighted sum of the set of the translated exemplar filters contained the usual filter bank. The arrangement allows for robust peak location detection yielding the target position and the multiplier angle that is extracted from the amplitude of the peak output response supplys an orientation estimate. A demonstration system using two such filters in an iterative fashion to counteract different sources of interference produced results accurate to plus or minus 1 degree 100 times faster than the traditional system.