By analyzing the behaviors of moving targets in video scenes, video violence detection can determine whether the targets' behaviors belong to violent behaviors such as assault, riot and fight.To detect violence in video, this paper proposed a video violence detection algorithm based on Gray Level Co-occurrence Matrix (GLCM) of optical flow field. The gray level co-occurrence matrix of optical flow obtains the spatial symbiosis distribution information of the motion in the scene and can calculate the important features to describe the behavior in the scene. Firstly, the optical flow of the video was extracted, Then, the Gray Level Co-occurrence Matrix of the grayscale image of the optical flow was extracted, the second moment, contrast, entropy and other features of the Gray Level Co-occurrence Matrix were calculated, and the feature vector was formed. Finally, support vector machines were used to classify the extracted video features into violent or non-violent behaviors. This method is applied to the three public data sets of Hockey, Movies and Violent Flow for 5 fold cross validation, and the accuracy rates are 96.7%, 95.8% and 92.5%, respectively.The experimental results show that the performance of the proposed method is superior to that of similar comparison method.