Formulate Feature Point Criterion For Estimation Telepointer Landmark
Nowadays, the growth of health care quality awareness leads to the advancement of medical technologies, especially surgeon technologies. In computer vision, tracking the tissues and internal organs (TDOD) movements has been beneficial to many surgical technologies. TDOD tracking poses a challenging...
Saved in:
| Main Author: | |
|---|---|
| Format: | Undergraduates Project Papers |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/39866/1/EA18011_MUHAMMAD%20AZIQ%20SYAUQI%20BIN%20HAMIDUN_THESIS%20-%20Aziq%20Syauqi.pdf http://umpir.ump.edu.my/id/eprint/39866/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Nowadays, the growth of health care quality awareness leads to the advancement of medical technologies, especially surgeon technologies. In computer vision, tracking the tissues and internal organs (TDOD) movements has been beneficial to many surgical technologies. TDOD tracking poses a challenging task due to the natural characteristic of TDOD, which mainly has a homogenous surface and texture. The main problem of tracking the tissues and internal organs (TDOD) movement are the homogenous texture of tissues and organs affect the image enhancement output and point detection since it is based on the corner of a bloodline. Next, using keypoint criteria will detect many points around the region. However, it contains not valuable points. Another problem is that it also limited the study of feature points criteria. Thus, this research objectively explores formulating the suitable feature point criterion to enhance the image for features extraction, identify features point criteria for estimating telepointer landmarks and validate the performance measurement by matching accuracy. The 5 types of image enhancement applied which is Contrast limited adaptive histogram equalization (CLAHE), Fast local laplacian filter (FLLF), Butterworth bandpass filter (BWBPF), Jerman enhancement filter, and Frangi filter. Its performance measurement is measured using lightness order error (LOE) to determine the naturalness preservation of enhancement. FLLF enhancement obtained the lowest value of LOE among the other type of image enhancement which for better naturalness preservation of enhancement. Then, the feature detection is computed to detect the number of keypoints of the enhanced image. Nine types of feature point detection is applied; SURF, SIFT, ORB, KAZE, Harris Corner Detection, BRISK, FAST, MinEigen and MSER. Therefore, KAZE point detection is only computed as the process of detection with the Jerman enhancement based on the analysis of the best selection. The features point criteria that approach are the area of ROI, number of point detection, point matching accuracy, and stableness of point detection. The estimating landmark process method is implemented with three methods; means, minimum point detection with template matching, and geometric transformation. All the feature point criteria successful meet the accuracy of estimating landmark. In findings, the performance output tested on three types of dataset; rotation, zoom in and out and liver deforming for the estimating landmark is 70% accurate. |
|---|