CAJ | 학술논문

基于视频内容植入广告系统目的是在用户观看视频时，能实时获取与视频内容相关商品的广告咨询。给出了系统解决方案，并分析了问题难点，提出了解决视频广告匹配问题难点的处理算法：先简单分割，然后通过直方图百分率方式二次匹配过滤后，得到分割后的图像，再利用改进的SIFT 算法进行精确匹配，从而确定当前视频图片中是否存在广告图片相似度高的商品信息。实验表明，基于视频内容的广告植入系统中图片匹配算法在常用视频中，算法实时性能好，匹配率高。
기우시빈내용식입엄고계통목적시재용호관간시빈시，능실시획취여시빈내용상관상품적엄고자순。급출료계통해결방안，병분석료문제난점，제출료해결시빈엄고필배문제난점적처리산법：선간단분할，연후통과직방도백분솔방식이차필배과려후，득도분할후적도상，재이용개진적SIFT 산법진행정학필배，종이학정당전시빈도편중시부존재엄고도편상사도고적상품신식。실험표명，기우시빈내용적엄고식입계통중도편필배산법재상용시빈중，산법실시성능호，필배솔고。
The purpose of video advertisement system is real-time obtainment of advertisement consultation of relative goods showed in video when users watching video. The contents of this research are the solution of system, analysis of difficult points, and solution of the difficult process on image matching with video ads. Firstly, it segments image simply. Then, it filters unmatched image twice by percentage of histogram method. At the last, it matches the image after segmenting precisely by simplified SIFT. We can receive the information whether the goods is the most similar to the image from video ads. Experiments show that this video advertisement system is real-time and possessing high matching rate on certain condition.