A weakly supervised framework for abnormal behavior detection and localization in crowded scenes

Positive unlabeled learning‐based anomaly detection in videos

“…Our method has competitive advantages in both data sets. TSR‐AE 11 and FRCNN 13 methods do not achieve the expected performance as the one‐class classification‐based methods considering the abnormalities that differ from the normal ones with potential misjudgment. Figure 4C,D show the ROC curves comparison on the pixel‐level criterion with Ped1 data set and Ped2 data set respectively, in which can be observed that our method achieves a better performance in anomaly detection.…”

“…The quantitative comparisons of AUC and EER metrics on the Ped1 data set are computed and listed in Table 2. Our method is a little superior to semi‐supervised learning methods, 11,13 one‐class classification methods, 1,16 and other PU learning methods 4,20 . For the Ped1 data set, we reduce 16.7% and 15.1% for the EER metric, and promote 20.2% and 23.1% for the AUC metric with TSR‐AE 11 in the frame level and pixel level, respectively.…”

“…(iv) Heuristic approaches , a supervised learning algorithm is first trained based on the labeled data only. The building classifier is then applied to the unlabeled data to generate more labeled examples as input for the supervised learning algorithm 13,14 …”

Multiobjects Association and Abnormal Behavior Detection for Massive Data Analysis in Multisensor Monitoring Network

“…e model proposed in this paper is mainly used in multisensor monitoring network to achieve multiobject detection and association and to analyze the behavior trend of the object. In order to evaluate the performance of our proposed abnormal behavior detection algorithm, the training set uses the international common behavior recognition data: UCSD, KTH，UCF101, and HMDB5 [31][32][33][34][35]. e KTH database includes six types of behaviors with 25 different pedestrians in four scenarios, namely, walking, running, waving, jogging, boxing, and handclapping, where the camera of these samples is relatively fixed and the background is simple; UCF101 and HMDB5 are complex datasets containing a large number of behavior categories, most of the data comes from video clips, pedestrian movement is complex, the angle of view changes greatly, and there are a lot of multiperson interaction.…”

“…e initial values of network weight W and bias b are set to random numbers with approximate 0, which are continuously adjusted by the Mathematical Problems in Engineering network during the training process to obtain the meaningful spatial information in the multisensor monitoring data. e adopted features are extracted from convolution layer and pooling layer, and then these features are sent to the long-term and short-term memory network units for analyzing the abnormal behavior [32]. e adaptive moment estimation (AME) algorithm based on stochastic gradient descent [28] is used to train the model.…”