Crime is the factor increases day by day and also needs the solution for identifying these activities in an efficient and quick manner. Many surveillance systems use artificial intelligence and image processing are incorporated with them to implement an intelligent surveillance system. But most of the systems are provided the alarm or identifies the crime after it happens. To solve this problem, camera footage-based theft detection will be used with the help of machine learning to detect theft occurrence. System will detect the human activity with the help of open-pose algorithm and convolution neural networks. Then the footage will be examined based on the pretrained model and it will be classified into three categories namely safe, abnormal or crime. Convolution neural network is used to classify the motion and an alert message will be sent to the owner along with captured image and options such as neglect or call the police.