<p>This research work proposes a novel SPOT-CRIME model for detecting and tracking suspicious objects and people from video footage to assist the crime scene analysis. The SPOT-CRIME model has three different monitoring phases for detecting and tracking suspicious object and person. The deep learning-based YOLO network is used in the Primary Monitoring Phase to identify potentially suspicious activities based on individual pose and motion estimates. A suspicious activity database is used for further analysis to detect the activities accurately. Secondary Monitoring Phase uses stochastic gradient Grad-CAM to identify objects related to individuals in the scene using suspicious object detection. Correlations between suspicious objects and activities detected in the first phase enable the detection of these objects. The Tertiary Monitoring Phase combines the outputs of the previous phases for tracking people and objects. In Track RCNN, individuals and suspicious objects are tracked robustly and associated with multiple frames. Detection accuracy of 98.54% was achieved by the proposed SPOT-CRIME model on benchmark datasets. Moreover, the performance of the SPOT-CRIME model progresses the overall accuracy range by 6.63, 1.05, 8.66 and 3.15% better than Seven-layered CNN, Enhanced CNN, Lightweight multiclass-CNN and CNN-BiLSTM respectively.</p>