Rahman, Aviv Yuniar (2017) Video Object Extraction Berbasis Lvq Menggunakan Metrik Jarak Minkowski Dan Euclidean. Masters thesis, Institut Teknologi Sepuluh Nopember.
|
Text
2215205012-Master Thesis.pdf - Published Version Download (3MB) | Preview |
Abstract
Minor stroke merupakan permasalahan penyakit utama di negara berkembang. Apabila penyakit penyakit minor stroke tidak segera diatasi akan berakibat lebih parah lagi. Deteksi penyakit minor stroke biasanya seperti Magnetic Resonance Imaging (MRI), histology, National Institutes of Health Stroke Scale score (NIHSS) dan Paroxysmal Atrial Fibrillation (PAF). Deteksi penyakit minor stroke memerlukan proses waktu dan tenaga. Padahal penyakit minor stroke harus segera ditangani. Supaya tidak berakibat pada kerusakan kognitif yang lebih parah, maka memerlukan sistem deteksi dan rehabiltas menggunakan video. Untuk tahap deteksi dan rehabilatas memerlukan proses salah satunya video object extraction. Penelitian mengenai video object extraction pada kasus minor stroke menggunakan LVQ telah dilakukan sebelumnya. Namun hasil akurasi maksimal 68.76% pada K=4.3. Kami mengusulkan perbaikan penelitian sebelumnya dengan mengganti merik euclidean dengan minkowski distance pada vector quantization (VQ). Serta mengukur kecepatan waktu dalam menyelesaikan ekstraksi pada metrik minkowski dan euclidean distance. Data yang yang dipergunakan menggunakan video orang terserang penyakit minor stroke. Untuk data pembanding menggunakan data video claire. Karena hanya memiliki satu video minor stroke saja. Metode yang dipergunakan dalam ekstraksi video minor stroke dan claire adalah learning vector quantization (LVQ). Video minor stroke dan claire diuji dengan variasi konstanta K=0.1 sampai K=5. Hasil yang diperoleh saat pengujian minor stroke dan claire dengan perbandingan metrik minkowski dan euclidean distance adalah akurasi sama sebesar 68.76% pada K=4.3. Hal ini dipengaruhi oleh kualitas video minor stroke kurang maksimal dan parameter konstanta ekstraksi fitur (K) dan konstanta metrik minkowski distance (P). Namun untuk hasil akurasi rata-rata pengujian claire extraction dengan minkowski distance lebih baik daripada metrik euclidean distance sebesar 72.49%. Sedangkan untuk hasil pengujian kecepatan waktu claire extraction dengan metrik minkowski distance lebih cepat 52 detik pada K=4.4 daripada euclidean distance. =============================================================================================== Minor stroke is the main illness in developed countries and should be prevented to avoid further severe injury. In order to prevent the illness, several detection methods have been developed, such as Magnetic Resonance Imaging (MRI), Histology, National Institutes of Health Stroke Scale score (NIHSS) and Paroxysmal Atrial Fibrillation (PAF). It is commonly known that minor stroke detection takes time and energy; thus, efficient video detection and rehabilitation method is required to be able to quickly detect the symptoms with a view to prevent the cognitive impairment. One of the processes in detection and rehabilitation is video object extraction. Some researches about video object extraction for minor stroke using LVQ has been conducted; however, the maximum accuracy achieved was 68.76% with K=4.3. We propose the use of minkowski distance instead of euclidean in vector quantization (VQ). Here, we measure the time to complete an extraction in minkowski and euclidean distance. Video data from patients with minor stroke is used and video data claire is used for comparison. With only one video minor stroke, a method to extract video minor stroke and claire is Learning Vector Quantization (LVQ). Video minor stroke and claire is tested with constant variant from K=0.1 to K=0.5. The same accuracy is derived from minor stroke and claire test with minkowski and euclidean matrix distance, namely 68.76% with K=4.3. This result is affected by a poor quality of video minor stroke, constant parameter extraction (K) and constant minkowski distance matrix (P). However, the mean accuracy for claire extraction with minkowski distance test is better than euclidean matrix, namely 72.49%. In addition the time of claire extraction with minkowski distance matrix is 52 seconds faster than euclidean distance with K=4.4.
| Item Type: | Thesis (Masters) |
|---|---|
| Additional Information: | RTE 621.388 33 Rah v-1 |
| Uncontrolled Keywords: | video object extraction, minor stroke, LVQ, minkowski distance, euclidean distance |
| Subjects: | T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques |
| Divisions: | Faculty of Industrial Technology > Electrical Engineering > 20101-(S2) Master Thesis |
| Depositing User: | Aviv Yuniar Rahman |
| Date Deposited: | 31 Oct 2017 03:56 |
| Last Modified: | 05 Mar 2019 03:32 |
| URI: | https://repository.its.ac.id/id/eprint/42402 |
Actions (login required)
 |
View Item |
