Object Recognition and Feature Detection Using MATLAB Sadhana

Object Recognition and Feature Detection Using MATLAB Sadhana Venkataraman 1, Yukai Tomsovic 2, Ms. Gangotree Chakma 3 Farragut High School 1, West High School 2, University of Tennessee Knoxville 3 1. TOPICS 1. 2. 3. 4. 5. 6. Introduction Edge Detection Shape Detection Color Segmentation Feature Detection Conclusion INTRODUCTION ● Method to perform operations on an image to get an enhanced image/extract useful information ● Basic Steps: ○ Acquire and import image ○ Analyze and manipulate image ○ Output -- image, component of image, or report based on analysis 2. EDGE DETECTION 3. SHAPE DETECTION ● ● Can be utilized to count objects and measure properties of individual objects ● Method: 1. Create a binary image (figure 2) 2. Label connected components from the binary matrix 3. Measure properties to classify objects Figure 1. Original image Figure 2. Binary image 4. COLOR SEGMENTATION ● Used to separate colors without clearly defined edges ● Method: 1. Convert RGB image to L*a*b color space to extract color information 2. Classify colors in ‘a*b*’ space with K-means clustering 3. Label pixels to separate clusters of color Method 1. Image with several small circles was created 2. Code used to find radii of all circles in terms of the number of pixels 3. Range was set between the radius of the smallest circle and the radius of the largest circle 4. MATLAB outlines the circles ● Practical uses ○ Astronomy -- classification of stars ○ Medicine -- detecting sickness ○ Archaeology -- finding objects Figure 3. Original Image Figure 4. Outlines of circles drawn by MATLAB ● Figure 6 -8: resulting segmented colors Applications of Color Segmentation: ● Distinguish different types of tissue, cells, tumor borders, etc. ● Used in medical thermography to detect breast cancer, tendon/ligament injuries, areas of poor circulation Figure 5. Original Image of watercolor paint Figure 6. First cluster of color Figure 7. Second cluster of color Figure 8. Third cluster of color 5. FEATURE DETECTION ● ● Use key features to determine whether two images are identical Figure 9 -14 - compare original image (9) to rescaled and rotated image (10) 1. The image is rotated and rescaled -- (MATLAB’s job - to restore the distorted image back to its original form) 2. Color-coded overlay of the original and distorted images and determined the number of matching features (11) 3. MATLAB identified that the images were the same based on an if statement written in the code ● Same process used to determine whether pictures are different (Figure 15 -19) Figure 9. Original image Figure 10. Distorted Image Figure 11. Overlay of matching features Figure 12. Comparison of original (left) & recovered (right) images Figure 17. MATLAB identifies these pictures as different Figure 13. MATLAB identifies these pictures as the same. Figure 14. Image 1 Figure 15. Image 2 Figure 16. Overlay of Images 1 and 2 6. CONCLUSION Our research is a foundation for more advanced technology. Some further applications of image processing are in the medical field, for PET scans and x-rays, for space image processing, such as the Hubble space telescope images, and for character recognition in scanning license plates and zip codes. REFERENCES Anbarjafari, Gholamreza. "Introduction to Image Processing. " Sisu@UT. University of Tartu, n. d. Web. 13 June 2016. This work was supported primarily by the ERC Program of the National Science Foundation and DOE under NSF Award Number EEC-1041877.