Unlock instant, AI-driven research and patent intelligence for your innovation.

An image accurate segmentation method fusing a deep learning network and a watershed algorithm

A technology of deep learning network and watershed algorithm, which is applied in image analysis, image data processing, calculation, etc., to achieve the effect of improving detection accuracy

Active Publication Date: 2019-06-14
ZHEJIANG UNIV
View PDF8 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0011] In order to solve the problem of accurate segmentation in the background technology, the present invention provides an image accurate segmentation method that integrates the deep learning network and the watershed algorithm, and fine-segments the area around the initial segmentation result, which can improve the image segmentation accuracy

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • An image accurate segmentation method fusing a deep learning network and a watershed algorithm
  • An image accurate segmentation method fusing a deep learning network and a watershed algorithm
  • An image accurate segmentation method fusing a deep learning network and a watershed algorithm

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0034] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0035] Embodiments of the present invention are as follows:

[0036] A video camera (in this example, DS-2CD3T20-I3) and a hard disk video recorder (in this example, ST4000VX000) are used to continuously shoot and record images of multiple pregnant sows.

[0037] Step 1: Select 1000 sow images of different scenes, time periods and shooting angles, use commonly used image processing software for image recognition to obtain the sow outline, and turn the area outside the sow outline on the image into black as a data set.

[0038] Step 2: Randomly select 124 images from the data set as the test set, and the remaining 876 images as the training set, use DeepLab for model training, and obtain the sow recognition model.

[0039] Step 3: Pending images such as figure 1 As shown, the undetermined image is recognized by the sow recognition model, and t...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses an image accurate segmentation method fusing a deep learning network and a watershed algorithm. The method is characterized by adopting a DeepLab recognition model to recognizethe to-be-determined image to obtain an initial segmentation image, adopting the watershed algorithm to segment the to-be-determined image to obtain a group of to-be-determined areas, multiplying thenumber of the to-be-determined areas by the initial segmentation image points, dividing the to-be-determined areas into to-be-determined object areas, or else removing the to-be-determined areas in the to-be-determined object areas. The method comprehensively utilizes the distance between the to-be-determined point and the to-be-measured substance center and the gray difference between the to-be-determined point and the foreground and background to judge the attribute of the equal point, and achieves the precise segmentation of the image. The characteristic that adjacent pixels with similar gray levels are partitioned by the watershed is utilized, the core area of the to-be-detected object is established by adopting a deep learning method, and the detection precision is improved.

Description

technical field [0001] The invention relates to a method for further improving image segmentation accuracy on the basis of prior art image segmentation, and in particular to an accurate image segmentation method that integrates a deep learning network and a watershed algorithm. Background technique [0002] Image segmentation is the process of dividing an image into multiple specific regions to represent different things, and it is an important step in target recognition. [0003] In the behavior detection of pigs, it is necessary to identify pigs from various backgrounds, realize the image segmentation of pigs, and lay the foundation for further behavior analysis. However, due to the existence of various facilities in the pig farm and the constant changes in lighting conditions, traditional image segmentation methods are prone to failure. [0004] In recent years, deep learning methods have been applied in image segmentation. [0005] FCN is the earliest classic model of ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06T7/12G06T7/187
Inventor 饶秀勤宋晨波张小敏高迎旺应义斌泮进明郑荣进
Owner ZHEJIANG UNIV