中文

Research Field

Professor Zhao's main research area is image processing, focusing on medical image segmentation, pattern recognition, industrial image detection and other branches of research. Medical image segmentation is mainly in the field of liver segmentation, abdominal multi-organ segmentation, liver vessel segmentation, retinal vessel segmentation, cell division, etc. Pattern recognition has made some achievements in the field of natural scene text detection, image tamper detection and so on. Industrial image detecting focuses on defect detection of metallic foams.


Medical image segmentation is the key technology of medical image processing and analysis. From medical research and clinical application perspective, image segmentation is the basis of medical image processing,medical diagnosis and target identification. It concerns a lot of fields such as object extraction, quantitative analysis, 3D reconstruction and so on. The purpose of medical image segmentation is to divide the original 2D or 3D image into multiple regions with different properties (such as gray scale, texture, etc.), and thus the regions of interest can be extracted, providing reliable basis for clinical diagnosis and pathological study.


Image segmentation is a process of dividing the images into several regions according to the similarity between one region and another region. However, the medical image background is pretty complex, with severe noise and blurring boundaries between organs, which obstructs the development of medical image segmentation. The laboratory, working for these difficulties, had made some achievements, and some of the related papers are as follows:


[1] Automatic segmentation for cell images based on bottleneck detection and ellipse fitting

[2] A New Approach to Segment Both Main and Peripheral Retinal Vessels Based on Gray-Voting and Gaussian Mixture Model

[3] Retinal vessels segmentation based on level set and region growing

[4] Retinal vessel enhancement based on multi-scale top-hat transformation and histogram fitting stretching

[5] Fuzzy C-means Clustering-based Multilayer Perceptron Neural Network for Liver CT Images Automatic Segmentation


Some of medical image segmentation results:

    

(1) Liver segmentation results  (2)3D liver vessels segmentation results  (3)Retinal vessel segmentation results  (4)Cell segmentation results    

                                                 

Pattern recognition refers to processing and analyzing numerical, textual and logical relationships that characterize various forms of phenomena to describe, identify, classify and interpret targets, which is important for information science and artificial intelligence. In general, pattern recognition is to classify and identify target objects with multiple techniques, including machine learning, artificial neural networks, deep learning, etc. The laboratory has focused on the natural scene text recognition and image tampering detection, and has achieved certain results. Some of the related papers are as follows:


[6] Scene text detection using adaptive color reduction, adjacent character model and hybrid verification strategy

[7] An automatic video text detection method based on BP-adaboost

[8] Tampered region detection of inpainting JPEG images

[9] Detection of tampered region for JPEG images by using mode-based first digit features

[10] Passive detection of copy-paste forgery between JPEG images

[11] Passive Detection of Paint-Doctored JPEG Images

[12] Passive Detection of Copy-paste Tampering for Digital Image Forensics


Some of pattern recognition results:

     

                (1)Natural scene text detection results                                              (2)Forged images                                  (3)Tampered area      

Industrial image detection, using machine vision instead of the human eye for the detection of industrial products, can effectively improve the quality of products and reduce product cost. In this part, the laboratory has mainly focused on the defects of metal foam images, and achieved certain results. Some of the related patents are as follows:


[1] A method for on-line detecting skip plating defect of continuous strip foam metal material

[2] A method for on-line automatic detection of cavity defect of porous metal material

[3] A method for position skip plating area of continuous strip of porous metal materials

[4] Method of determining a continuous strip of porous metal material skip plating defects

[5] A method for detecting and position skip plating area of continuous strip of porous metal materials


Foam metal image defect partial detection results show:

  

      (1)Hole detection results                        (2)Crack detection results

 
  • ZHAO Yuqian
    • Zip Code:

    • Postal Address:

    • Mobile:

    • Email:

Central South University  All rights reserved  湘ICP备05005659号-1 Click:
  MOBILE Version

The Last Update Time:..