基于商标匹配的视频广告识别

视频广告作为一种重要的媒体形式，正在潜移默化地影响着人们的生活。针对如何对视频广告进行识别，本文提出了一种基于商标匹配的视频广告识别算法。考虑到视频广告帧中存在着大量的背景干扰问题，本文算法首先对视频广告帧进行显著区域检测，将广告帧中较为显著的商标信息与周围背景分开，然后对已得到的显著区域图进行SURF特征提取，并与商标库中的商标显著区域图进行匹配。实验结果表明:本文算法对广告帧中商标存在的背景干扰、尺度缩放、旋转、颜色、光照等变化具有很好的鲁棒性，取得了令人满意的匹配准确率，很好地实现了视频广告识别。 

Video Commercial Recognition Based on Logo Matching

Video commercial has been an important dissemination medium and been around every corner in human’s life presently. This paper proposes a video commercial recognition algorithm based on logo matching and do some research on the matching accuracy. Considering that the commercial logo is generally embedded in a video frame with complex and even cluttered background, the visual saliency detection is first applied to highlight product logo region, as well as suppress the influence from background on further commercial recognition. Then, the Speeded-UP Robust Feature (SURF) based image matching is implemented to identify whether one logo in the commercial logo database, which was built through the Web Spider crawling, appears in the video frame or not. The experimental results show that the proposed algorithm is exceedingly robust to some variations such as scale, rotation, color and illumination, which leads to a preferable matching accuracy and achieve a satisfactory video commercial recognition performance.