COMPRESSED DOMAIN VIDEO UNDERSTANDING METHODS FOR TRAFFIC SURVEILLANCE APPLICATIONS

In the realm of traffic monitoring, efficient video analysis is paramount yet challenging due to intensive computational demands. This thesis addresses this issue by introducing novel methods to operate in the compressed domain. Four methods are proposed for image reconstruction from High Efficiency Video Coding (HEVC) Intra bitstreams, namely, the Block Partition Based Method (Mbp the Prediction Unit Based Method (Mpu), the Random Perturbation Based Method (Mrp), and the Luma based method (My). These methods aim to provide a compact representation of the original image while retaining relevant information for video understanding tasks. Our methods substantially reduce data transmission requirements and memory footprint. Specifically, images created via Mbp and Mpu require 1/1,536 and 1/192 of the memory needed by pixel domain images, respectively. Moreover, these methods offer computational speedup between 1.25 to 4 times, yielding efficiencies in video analysis. The proposed techniques have shown promising results in detecting vehicle license plates and classifying vehicles, with accuracies matching or surpassing pixel area levels. For instance, Mbp and Mpu accurately identified license plate locations, achieving accuracies of 93.33% and 99.02%, respectively. The vehicle detection process showed speed enhancements and high accuracies with both Mrp and My methods. Our work sets the foundation for future research, exploring avenues such as full video decoding, hybrid approaches combining compressed and pixel domain information, and extending the proposed methods to other object detection applications. In conclusion, our methods present a significant stride towards efficient video analysis in traffic surveillance, enhancing computational performance while maintaining high detection and classification accuracy.