The demand for greater security measures for monitoring and protecting operations has made video anomaly detection one of the most significant study areas in the field of computer vision. This is due to the fact that the detection of anomalies in video surveillance systems has increased, making it one of the leading focus areas in the current field of research. When looking for suspicious behaviours like aggression, robbery, and wrong U-turns, assigning real people to frequently analyse the surveillance footage is a process that is both time-consuming and prone to error. As a direct consequence of this, there is an urgent requirement for the development of advanced, fully automated video surveillance systems for use in public areas. The research uses pipelined deep encoders to find a solution to the problem of locating and identifying anomalies in surveillance videos. The authors especially combined the LSTM autoencoder with the convolutional autoencoder. This strategy helps gather spatial and temporal information from the input video stream. This allowed us to achieve optimal results. The authors used the one-class classification principle during the training phase of the model. The training is done on normal data, and when they were verifying or testing it, they used anomalous testing data. The analysis of the study is based on standard benchmark practises for error rate and the duration of time required for identifying anomalies in the sequence of the video stream. This study satisfies the requirements for operating in near real time.