| Barcode |
|
| Judul English |
|
| Examiners |
|
| Head of post graduat |
|
| Tim penguji 3 |
| Anto Satrio Nugroho |
| No. Induk |
|
| Tim penguji 4 |
| Hisar Maruli Manurung |
| Kata Kunci |
| tracking, particle filter, prediction, update/ connection, human movement analysis |
| Pembimbing 3 |
|
| Pembimbing 2 |
|
| Kopromotor |
| Muhammad Rahmat Widyanto |
| Tahun buku |
| 2010 |
| Barcode RFID baru |
| 11636770 |
| Tim penguji 1 |
| Belawai H. Widjaja |
| Promotor |
| T. Basaruddin |
| Abstrak Indonesia |
| Saat ini berbagai macam aplikasi membutuhkan proses analisis pergerakan manusia dari data citra dinamis, antara lain, sistem pengawasan cerdas, interaksi manusia computer, analisis kinerja atletik, dan lain-lain. Analisis pergerakan manusia dari data citra dinamis, terdiri dari dua tahap utama, yaitu pelacakan dan analisis pergerakan manusia. Metode Particle Filter (PF) biasa digunakan pada tahap pelacakan pergerakan manusia. Metode ini mempunyai tingkat akurasi pelacakan yang tinggi. Hanya saja metode ini mempunyai dua kelemahan. Pertama, jumlah partikel yang dibangkitkan sebanding dengan waktu komputasi pelacakan. Kedua, banyak partikel sama yang dibangkitkan, sehingga waktu komputasi banyak terbuang untuk memproses partikel yang sama ini.Pada penelitian ini diusulkan modifikasi metode PF untuk mengatasi permasalahan tersebut. Metode yang diusulkan adalah metode PF dengan pembobotan Gaussian. Metode ini terdiri dari empat tahap, yaitu prediksi, sampling partikel, pembobotan Gaussian, dan koreksi. Prediksi bertujuan untuk membangkitkan sejumlah partikel yang merupakan representasi lokasi target obyek. Sampling partikel bertujuan untuk menghitung bobot partikel-partikel tertentu yang menjadi parameter-parameter Gaussian. Pembobotan Gaussian bertujuan untuk menghitung nilai bobot pertikel-partikel berdasarkan distribusi Gaussian yang terbentuk. Koreksi bertujuan untuk memperbaharui partikel-partikel berdasarkan bobot masing-masing. Dengan dua tahap tambahan yang merupakan modifikasi metode PF, maka perhitungan bobot semua partikel hasil prediksi tidak diperlukan lagi, sehingga dapat mengurangi waktu komputasi.
|
| Cat. Umum |
|
| Judul |
| Sistem analisis pergerakan manusia menggunakan metode particle filter dengan pembobotan Gaussian dan kombinasi sistem inferensi Fuzzy dan ukuran fuzzy |
| Co-Supervisor |
|
| Subjek |
| Fuzzy measure theory |
| Pembimbing 1 |
|
| Examiners 6 |
|
| Examiners 5 |
|
| Examiners 4 |
|
| Supervisor |
|
| Examiners 3 |
|
| Examiners 2 |
|
| Examiners 1 |
|
| Bibliografi |
|
| Pengarang |
| Indah Agustien Siradjuddin; |
| Co-Supervisor 1 |
|
| Cat. Karya |
|
| Tim Penguji 6 |
|
| Timpenguji 2 |
| Aniati Murni Arymurthy |
| Tim Penguji 7 |
|
| Tim Penguji 5 |
| Petrus Mursanto |
| Co promotors |
|
| chair Person |
|
| Tanggal Datang |
| 04/03/2010 |
| Asal |
|
| Kopromotor 1 |
|
| NPM |
| '0606038420 |
| Abstrak English |
| Currently human movement analysis from dynamic image is required in various system applications, e.g. surveillance system, human computer interaction, athletic performance analysis. Human movement analysis from dynamic image consists of two main stages, i.e., human tracking and human movement analysis. Commonly Particle Filter (PF) method is used in the human tracking process becuase of its advantages, for instance, high accuracy rate and simple algorithm. Unfortunately, this method has two main drawbacks. First, number of generated particles is proportional to computational time of tracking. Second, many same particles are generated, therefore there are wasted computational time to process the same particles. Modification of PF method is proposed in this research to cope with particle filter's drawbacks. The proposed method is named PF with Gaussian weighting. This method consists of four main stages, i.e., prediction, particles sampling, Gaussian weighting, and update/ correction. The objective of prediction is to generate a number of particles which represent target's location. THe objectiveof particles sampling is to obtain certain particles which represent parameters of Gaussian distribution. The objective of Gaussian weighting is to calculate particle weight based on Gaussian distribution. The objective of update / correction is to update particles based on each weight. The modified PF could reduce computational time of object tracking since this method does not have to calculate particle weight one by one. To calculate weight, the proposed method builds Gaussian distribution and clculates particle weight using this distribution. The porposed method has been tested on video data with various backgrounds and a number of subjects. The result of experiments shows that the proposed method reduces computational time in tracking process maximum 68% compared to the conventional one, meanwhile the accuracy of tracking with this new method is 88.69% for the artificial data video. For the benchmark data video, the proposed method could reduces computational time 68% compared to the conventional one and the tracking accuracy is 82.19%. Therefore the proposed method is promising for human movement analysis application, since it can be applied in real time application and has high accuracy rate. |
| Pengarang 2 |
|
| Chair of examiner |
|
| Fisik |
| xix + 158 pages;il. +figures, tables, attachment |
| Bahasa |
| ind |
| Lulus Semester |
|
| Penerbitan |
| Depok: Fakultas Ilmu Komputer UI, 2010 |
| No. Panggil |
| DIS-022 (Softcopy DIS-013) Source Code DIS-006 |