An improved line extraction method based on edge point grouping

Abstract

The invention provides an improved straight line extraction method based on edge point grouping, which relates to the technical field of image processing and analysis. The invention obtains a binary edge image I2 after preprocessing the image I to be processed, and defines 8 neighborhoods of the current pixel point image. The points of the first row, the first column, the last row and the last column in the image I2 do not participate in subsequent operations; Starting from the second row of I2, scanning from left to right, obtaining a data point set i, where i=1, 2,..., n; Hough transform iscarried out on that data point set i to obtain a straight line set BL to be merged and a straight line extraction result CL by judging parameters Rho and Theta; Further judging whether the straight lines in the set BL should be merged; the obtained set C is that final straight line extraction result. The method improves the grouping efficiency of the edge points, greatly reduces the time consumingof the straight line extraction process, and has higher straight line extraction efficiency.

Description

technical field [0001] The invention relates to the technical field of image processing and analysis, in particular to an improved line extraction method based on edge point grouping. Background technique [0002] Straight lines are important geometric features in images, which can accurately describe the shapes of man-made objects such as roads and houses, and are widely used in data processing processes such as image matching and 3D reconstruction. Therefore, line extraction and matching has always been a research hotspot in the field of photogrammetry and computer vision. Among the classical straight line extraction methods, Hough transform straight line extraction obtains a small number of straight lines, but the program takes less time. The straight line extracted by the LSD (Line Segment Detector) method is the best, but due to the limitation of the threshold and the selection of the seed point, some straight line segments are missed, and the program is relatively tim...

AI Technical Summary

Problems solved by technology

The straight line extracted by the LSD (Line Segment Detector) method is the best, but due to the limitation of the threshold and the selection of the seed point, some straight line segments are missed, and the program takes a relatively long time

[0003] In 2013, a phase-grouped line extraction method based on adaptive partitioning was proposed. According to the principle of the maximum number of pixels in the support area, the line feature extraction is realized, but the continuity of the extracted line still has room for improvement.

[0004] In 2013, an improved LSD algorithm based on iterative verification was proposed, that is, the verification process of the support domain was improved by changing the maximum tolerance of the gradient direction angle of the linear support domain and the radius R of the linear support domain, but more comparisons of pixel gradient directions were required. , it is easy to lose tiny details; in addition, an edge image-based LSD line extraction algorithm is proposed, firstly calculate the gradient modulus map and direction map, then extract the edge support points of the coefficients from the modulus map, and finally connect these points through path planning The support point extracts the edge, although the integrity and continuity of the detection results are improved, but the image noise is amplified, and many unnecessary tiny details are detected

[0005] In 2015, a method of combining LSD and Mean Shift algorithm was proposed to extract straight lines. This algorithm uses MeanShift to solve the problem of extracting significant straight lines, but there will still be a small amount of false straight lines in the result.

[0006] In 2016, the edge detection step was added to the traditional LSD algorithm, and the approximate area and similar straight line were merged to encode the three-dimensional combined target. However, when merging straight lines, the algorithm easily extracts the closer straight lines as a straight line.

[0007] In 2017, the straight line fitting angle prior method was added to the Hough transform, which solved the problem that the Hough transform generated a large number of short straight lines and arcs. However, when extracting buildings, small rectangles and east-west straight lines will be missed.

[0008] Although the above methods have achieved certain results, they all have the disadvantages of less linear results extracted by some programs or more time-consuming programs.

Images