X-ray scanners are increasingly used for scanning vehicles crossing international borders or entering critical infrastructure installations. The ability to penetrate through steel and other opaque materials and the nondestructive nature of x-ray radiation make them ideal for finding drugs, explosives and other contraband. In many situations, the same vehicles cross the checkpoint repeatedly, such as the employee vehicles entering a high-risk facility or cargo vehicles crossing international borders back and forth. Manual analysis of these images puts extra burden on the operator and results in slow throughput. In this paper we report an integrated and fully automated system to solve this problem. In the first stage of the algorithm, a model-based segmentation approach is used to find the vehicle outline. It proceeds by first using background subtraction to find the overall body of the vehicle. Next, we find the outlines of tires by using rotating edge detection kernels. The lower outline of the vehicle is found using active contours. We then use a deformable registration approach to align the vehicles which is specifically designed for the requirements of this problem. An intensity normalization step is then performed to account for the intensity variations between the scans at two time points. We use a histogram-based approach that scales and shifts the histogram of one image to match that of the other. The differences between the two inspection results are computed next. We then apply knowledge-based rules to remove false alarms such as lights and driver's body. The system is specifically designed for back-scatter x-ray imaging which is a powerful modality for detecting organic materials such as drugs and explosives. We have applied this system to images scanned by a deployed x-ray scanner and have achieved satisfactory results.

Detecting regions of change in multiple images of the same scene taken at different times is of widespread interest due to a large number of applications in diverse disciplines, including remote sensing, surveillance, medical diagnosis and treatment, civil infrastructure, and underwater sensing. This paper presents a systematic survey of the common processing steps and core decision rules in modern change detection algorithms, including significance and hypothesis testing, predictive models, the shading model, and background modeling. We also discuss important preprocessing methods, approaches to enforcing the consistency of the change mask, and principles for evaluating and comparing the performance of change detection algorithms. It is hoped that our classification of algorithms into a relatively small number of categories will provide useful guidance to the algorithm designer.