An Innovative Method for Snatch Theft Detection Using Convolutional LSTM-Auto Encoder Approach

Urban areas frequently experience snatch stealing, a severe threat to public safety. It entails forcibly seizing a victim's personal items, including purses or cell phones, before scurrying away from the scene. Due to how quickly they happen, it is difficult to identify snatch theft instances in real-time. The systems now in use to identify the snatching occurrences mainly rely on human participation, which may result in lengthy delays and mistakes. Consequently, there is a need for an automated method that can quickly and effectively identify these situations. Preprocessing, segmentation, feature extraction, and model training are all used in the suggested method. Gaussian Model is utilized for preprocessing. Canny edge is utilized for segmentation. LSTM-AE is used to train the model. Two further models, LSTM and AE, are contrasted with the suggested method. When compared to two existing approaches, the proposed strategy performs exceptionally well.