Characterization of Stones and Logs in a Chipper Log Feed using 3D Image Analysis
Independent thesis Advanced level (professional degree), 20 credits / 30 HE creditsStudent thesis
In this thesis a solution for finding stones mixed up with logs in the timber industry is presented. The main goal is to identify as many stones as possible without giving false alarms when there are no stones. The solution uses merely the surface height profile obtained from a stereo vision setup above the conveyor belt for identifying the stones. A 3D-image with objects heights and defined axis can be obtained after camera calibrations and be used to determine the size of objects.Two separate methods for finding stones in 3D-images has been developed. Stones must be at least 5 cm high, be fairly circular with a radius of at least 6 cm and be clearly visible in the stereo images.The first method is an experimentally determined approach that has been developed to detect small stones and large stones separately. Vertical logs are morphologically reconstructed and removed leaving only possible stones.The second method uses a frame of the 3D-image as a mask for reconstructing and identifying logs no-matter the direction of the logs. The logs are removed and if there is a possible stone with a high enough ratio of how well the circular structuring element describes the stone it is characterized as a stone.Both methods gives no false identified stones in 1297 images without stones and the first method has a hit-rate of 60% while the second method has a hit-rate of 67% in 120 images containing stones. My conclusion is that the methods gives equivalent results but I would prefer the second method since it is simpler, has less adjustable values and is faster computationally. The results suggests a robust algorithm and the possibility to identify stone merely by the surface height profile.
Place, publisher, year, edition, pages
2012. , 96 p.
IdentifiersURN: urn:nbn:se:ltu:diva-56955Local ID: dae1a2e1-642b-48ac-b7e9-a3fd7f8146b5OAI: oai:DiVA.org:ltu-56955DiVA: diva2:1030342
Subject / course
Student thesis, at least 30 credits
Engineering Physics and Electrical Engineering, master's level
Thurley, MatthewOlevik, David
Validerat; 20120812 (anonymous)2016-10-042016-10-04Bibliographically approved