Tree annual ring automatic detection method based on deep learning

Abstract

The invention belongs to the technical field of tree annual ring detection, and particularly relates to a tree annual ring automatic detection method based on deep learning. The method comprises the following steps: performing data acquisition; deblurring the image; performing data normalization: carrying out MinMax normalization on the data; dividing data; expanding data; constructing a model; storing the model when the loss function of the model does not drop any more and the evaluation index is optimal and tends to be stable. According to the method, image expansion is carried out by usinga synthesis method, the problems of model overfitting and difficult improvement of detection performance due to too small data volume are solved, instance segmentation errors caused by coordinate deviation are eliminated by independently carrying out Mask prediction in the Mask-RCNN model, and the detection performance of the model is improved. The construction of the model can solve the problem that an existing automatic method still needs a large number of users to interact, accelerate the detection time of the larch annual ring disc and improve the detection efficiency. The method is used for detecting the tree ring.

Description

technical field [0001] The invention belongs to the technical field of tree ring detection, and in particular relates to an automatic tree ring detection method based on deep learning. Background technique [0002] Tree ring width is an important source of climate and historical data, providing an important data source for dendrochronologists, climatologists, archaeologists, etc. But measuring these widths often requires a lot of manual work. [0003] Problems or deficiencies with existing techniques: Considerable progress has been made in the automation of this task in computer vision and machine learning, but most automated methods still require extensive user interaction, such as marking centers or measuring paths. Also deformation of the rings in the tree rings, double or missing rings, and cuts in the wood make reliable identification extremely difficult. Contents of the invention [0004] Aiming at the technical problem that the above-mentioned automatic method req...

AI Technical Summary

Problems solved by technology

[0003] Problems or deficiencies with existing techniques: Considerable progress has been made in automating this task in computer vision and machine learning, but most automation methods still require extensive user interaction, such as marking centers or measuring paths

Also deformation of the rings in the tree rings, double or missing rings, and cuts in the wood make reliable identification extremely difficult

Images