Study On Image Detection Approach Of Tree-Ring Parameters

Tree-ring width is one of the conventional research objects in dendrochronology, detected by manual to the present; it depends heavily on the operator's surveying experience and operational level, and is very prone to eye-tiredness. As is considered to be more correlative to environment factors, tree-ring area has been increasingly admitted by scholars recently, but it can not be measured directly by traditional methods, instead, it just estimate value based on tree-ring width. Aiming at this problem, this paper takes larch and robinia pseudoacacia as studying objects, and use computer vision technology to detect tree-ring automatically. The main contributions of this paper as follows:(1) According to physical meanings and tree-ring image features, a series of automatic identification algorithm including tree-ring image processing and parameters detecting was designed. The flow could also applicable to other species.(2) To meet the grey difference between latewood and heartwood, proposed that the key step of parameter detecting was to extract latewood boundary. Proposed a dynamic threshold segment algorithm based on regional sub-block shape. As boundary shape of tree-ring heartwood and sapwood approximates to round, ellipse boundary was applied to block image. Boundary ellipse was determined by variant features of the histogram. This method can simulate boundary shape of core and sapwood better, and saved tree-ring boundary which has distinct diversity in ground grey primely, which conquered touching problems between blocks existing in traditional matrix block method. In view of the difference between object and background in perimeter, area etc. geometrical parameters, brought out a contour segment method on the basis of region geometrical features, which can remove splashes in larch ring images and holes in locust tree images efficiently.(3) After considered their own physical meaning and the conventional survey steps fully in recognize ring parameters, proposed measurement approach of ring width and area based on contour tracing. This method was not affected by internal holes of latewood, and was more accurate and showed strong adaptability, compared with common pixels statistical method(4) Computer vision-based tree-ring parameter detecting system was developed by VC++ 6.0, which has friend interface and was very convenient in processing and storing on detecting results. The system could recognize parameters automatically, and manual...