A Watershed Segmentation Method Combined with Improved Fast Merging Algorithm

Abstract

The invention discloses a watershed segmentation method combined with an improved fast merging algorithm. The method comprises the following steps: step 1) research and develop a set of pre-processing based watershed segmentation algorithm technical process based on GF‑1 data; step 2) research based on the previous In the processed watershed segmentation algorithm, the influence of the global marker threshold α and the regional adaptive threshold adjustment coefficient g on the segmentation results is analyzed, and the value ranges of α and g are analyzed; step 3) based on the initial segmentation results of the pre-processing watershed segmentation algorithm, the RAG and NNG as the implementation method, use the Fast Lambda merging algorithm based on the controllable common boundary length penalty to merge the initial segmentation results to achieve a satisfactory segmentation result; step 4) research the Fast Lambda merging algorithm based on the controllable common boundary length penalty The influence of the threshold value T and the public boundary penalty coefficient λ on the merging result of the control area merging is analyzed, and the value range of T and λ is analyzed.

Description

technical field [0001] The present invention relates to a watershed segmentation method combined with an improved fast merging algorithm, especially using two image segmentation algorithms (watershed segmentation algorithm based on pre-processing and Fast Lambda algorithm based on controllable common boundary length penalty) for high-resolution remote sensing images Perform image segmentation to obtain good image segmentation results. Background technique [0002] In remote sensing application analysis, remote sensing image segmentation is a bridge between low-level image processing and middle-level image analysis and understanding, and is a key step in completing the precise classification of high-resolution remote sensing images. In general, image segmentation algorithms include: threshold segmentation algorithm, edge segmentation algorithm and region segmentation algorithm. [0003] For the image threshold segmentation algorithm, this method has a small amount of calcula...

AI Technical Summary

Problems solved by technology

[0009] Aiming at the problems of over-segmentation and loss of contour information when using the watershed algorithm for high-resolution remote sensing image segmentation, the present invention combines the GF-1 image and utilizes object-oriented technology to establish a fast lambda combination based on the controllable common boundary length penalty. Algorithmic Watershed Improved Segmentation Method

Images