Elliptical object segmentation method based on ellipse fitting

Abstract

The invention discloses an elliptical object segmentation method based on ellipse fitting, and the method comprises the steps: obtaining an object of a target region in an image through extracting a maximum connected region of a G component value in an image RGB gradient value, removing the interference of background factors in the image through employing an RGB ultra-green segmentation algorithm, and removing a non-target region through employing a top-hat transformation method; extracting a central point of the elliptical object by combining a rapid radial symmetry method based on distance transformation and a global threshold value; segmenting the edge part of the elliptical object by using a concave point detection and contour estimation method; and finally, complementing the shielded parts of the overlapped objects by using an ellipse fitting algorithm based on least squares. According to the method, accurate extraction of overlapped target edge points is ensured; the problems of incomplete object edge extraction and non-ideal segmentation effect caused by similar RGB color gradient values in the image are avoided; the problem that the complete contour of the partially overlapped elliptical target cannot be obtained in the two-dimensional image under the view angle of the camera is solved.

Description

technical field [0001] The invention relates to the technical field of image segmentation, in particular to an ellipse-shaped object segmentation method based on ellipse fitting. Background technique [0002] Machine vision is a hot research topic of artificial intelligence. In machine vision research, object segmentation and parameter calculation based on images is a key issue. Object detection and segmentation in images also has many applications in reality, such as medical diagnosis, quality inspection of industrial parts, fruit sorting, etc. [0003] A lot of research has been done around object detection and segmentation in images. The commonly used methods to segment the target object from the background image are: edge detection method, threshold segmentation method and morphological segmentation method. In addition, some researchers use the improved watershed segmentation algorithm to extract the surface leaves in the image. The specific method is to first use the...

AI Technical Summary

Problems solved by technology

[0004] In the existing overlapping object segmentation methods, one disadvantage is that the edges of objects with similar RGB color gradient values ​​in the image cannot be extracted well, resulting in object segmentation results that do not achieve the desired effect; another disadvantage is that the occluded objects cannot Completing the missing part of the target object, resulting in inaccurate subsequent calculation results of the target object

When using deep learning methods to segment overlapping objects, not only the above two problems will occur, but also problems such as excessive labor costs will occur due to the need to manually label the data set in advance.

When using the convex hull filling algorithm to complement the occluded parts of overlapping objects, it is considered that the edge points at the overlapping objects in the image will deviate, so the algorithm cannot guarantee the correct prediction of the missing parts of such objects

Images