This paper presents a system for detection and recognition of pests in stored-grain based on video analysis. Unlike current systems which conduct analysis of static images, the proposed system uses video data captured by camera and performs video analysis to detect and recognize pests in grain. By using video data instead of static images, techniques such as motion estimation and multiple-frame verification are used to locate, count and recognize pests. Compared to systems based on image processing, the proposed system is more robust to moving pests and avoids missing and re-counting of moving pests. Furthermore, by analyzing motion of pests in video, the system can only count living pests and ignore dead ones, which are recommended by national standard of grain quality and cannot be achieved by current systems based on static image processing.