With global trade highly increasing, Non-Intrusive Inspection (NII) of containerised cargo using X-ray scanners has been identified as one of the important areas of the World Customs Organization (WCO) SAFE Framework of Standards. The framework aims to promote the secure movement of goods through international trade supply chains. Non-Intrusive inspection systems, such as X-ray instruments, have become integral in imaging the contents of cargo containers, serving as a primary tool for identifying potential threats, contraband and mis-declared items. These scanned images are analyzed alongside shipping manifest data to efficiently target containers for thorough physical examination. Considering the bottlenecks in manual analysis of scanned images, AI based Image analysis for automated risk detection is emerging as critical area in Non-intrusive inspection (NII). With paradigm shift and rapid advancements in AI based Image analytics from Image classification towards Object detection algorithms, this paper explores the potential of deep learning to enhance NII by employing cutting edge object detection algorithms for automated threat detection. This paper evaluates the performance of four SOTA deep learning architectures Faster R-CNN, Retina-Net, YOLOv5, and YOLOv7 on a custom Scanned X-ray Image dataset of NCTC (National Customs Targeting Centre) of Indian Customs. Our findings demonstrate the superiority of YOLOv7, achieving a mean Average Precision (mAP) of 0.9144, highlighting its effectiveness in object detection within X-ray images of containerised cargo. Beyond mAP, the paper demonstrates the importance of considering factors like higher inference speed and computational efficiency for real-world deployments. Additionally, this research contributes to the advancement of concept of Autonomous image analytics based Risk management framework for Customs administrations and signifies the pivotal role of AI and open source algorithms in bolstering border security.