vi
ABSTRAK
Temu balik video ataupun temu balik video berdasarkan konten gambar adalah salah satu bidang penelitian dari sistem temu balik informasi. Content Based Video Retrieval memiliki empat tahapan yaitu segmentasi video yang menghasilkan frame – frame, ekstraksi keyframe merupakan tahapan memilih frame kunci dari semua frame koleksi video, tahapan ekstraksi fitur menggunakan algoritma Speeded-Up Robust Features (SURF). Surf adalah detektor dan deskriptor fitur citra yang cepat. Dan kini banyak digunakan secara luas dalam aplikasi kompter visual seperti dalam aplikasi temu balik video.Surf pertama kali dipresentasikan oleh Herbert Bay pada konferensi Eropa 2006 tentang Computer Vision dan sebagian terinspirasi oleh Scale-Invariant Features Transform (SIFT). Titik interest atau key point dari citra query dan keyframe direpresentasikan oleh vektor yang kemudian akan dibandingkan dalam tahap pencocokan fitur. Tahap pencocokan fitur dilakukan dengan membandingkan nilai fitur pada queryyang diberikan dengan keyframe pada koleksi video setiap kategori. Parameter yang digunakan untuk mengukur kualitas dari hasiltemu balik video adalah nilai dari Recall, Precision, dan Running Time. Berdasarkan hasil pengujian menggunakan 2 jenis query, Query Bukan Frame yaitu query yang diambil diluar frame koleksi video dan Query Dari Frame yang diambil dari frame koleksi video.rata rata nilai Recall54,75%, nilai rata – rata Precision 37,5%, serta nilai Running Time 73,56 detik. Query Dari Frame dengan nilai rata – rata Recall 51%, nilai rata – rata Precision 59%, dan Running Time 121,67 detik.
Kata Kunci : Video, Citra, Temu Balik, Ekstraksi Fitur, Content Based Video Retrieval (CBVR), Speeded-Up Robust Features (SURF).
vii
IMPLEMENTATION OF CONTENT BASED VIDEO RETRIEVAL WITH
SPEEDED-UP ROBUST FEATURES (SURF)
ABSTRACT
Image Retrieval (IR) or Content-Based Image Retrieval (CBIR) is a well-known field of research in Information Retrieval System (IRS). Content Based Video Retrieval has four stages: video segmentation that produces frames, keyframe extraction is the stage of selecting key frames of all video frames, feature extraction stages using Speeded-Up Robust Features (SURF) algorithm. Surf is a fast image detector and descriptor feature. It is now widely used in visual compiler applications such as in video.Surf backlash apps were first presented by Herbert Bay at the 2006 European conference on Computer Vision and partly inspired by the Scale-Invariant Features Transform (SIFT). The interest point or key point of the query image and keyframe is represented by the vector which will then be compared in the feature matching stage. The feature matching stage is performed by comparing the feature value of the given query with the keyframe on the video collection of each category. The parameters used to measure the quality of video feedbacks are the values of Recall, Precision, and Running Time. Based on the test results using two types of queries, Query Not Frame is a query that is taken outside the frame of the video collection and Query From Frame taken from frame collection video.Average Recall value 54.75%, Precision average 37.5%, and Running Time value 73.56 seconds. Query From Frame with Recall 51% average value, Precision 59% average value, and Running Time 121.67 sec.
Keywords : Video, Image, Retrieval, Feature Extraction, Content Based Video Retrieval (CBVR), Speeded-Up Robust Features (SURF).