Bag Of computational verbs (BoCV) is a new framework based on computational verb theory. In this framework, verb similarities are summed up or averaged. The value of result is put into a bag as an entry for description in order that many verbs similarities can make up the final feature vector implicitly. A novel model called spatiotemporal verb bag (SVB) is proposed and it is trained in a supervised learning manner for two classes classification problem. While the application of this framework for the same problem is not confined to the proposed model. We compare the classification result of the proposed model with some baseline methods, e.g. histogram of oriented gradient feature with adaptive boosting (HOG-Adaboost) and principal component analysis with support vector machine (PCA-SVM). The proposed algorithm achieves excellent performance in distinguishing Chinese license plates captured from different parts of the day and variety of weather condition.