The teaching recording and broadcasting system based on image recognition technology usually requires the close-up of the indoor students from the camera by automatic identification and position. A head detection algorithm based on multi-feature fusion for the indoor students is proposed to solve this problem. Firstly, head candidate regions are extracted through setting a threshold value based on the color characteristics of human hair. Then the head region of all the students can be detected by shape features of the hair region and skin color features of the face region. Finally, the speaking student can be captured by detecting the position change of the center of the contour in the vertical direction of the image when he stands up. The method was used for image processing of several angles of classroom video, and the experiment results show the effectiveness of it in detecting the head region of students.