Edge Intelligence with Long-distance Micro Objection Detection and Low-light Image Enhancement for Oil Gas Pipeline Monitoring

Abstract

The smooth operation of the oil and gas pipeline network plays a vital role in ensuring the safety and stability of the national energy supply. The implementation of pipeline protection is a necessary measure to ensure the smooth operation of the pipeline network, and video surveillance is an important technical means of pipeline protection. Existing video surveillance technology has the following problems in longdistance micro-targets, low-light images and intelligent identification in harsh environments. To tackle these problems, in this paper, we propose a set of optimization techniques to improve the situation. First, we propose a new objection detection method that can handle long-distance micro-targets. Here, the object detection refer to the detection of moving targets such as vehicles, people, animals. Second, we propose a transfer learning based photo-enhance technique under low light. We have implemented our techniques on RK3399 platform and extensively verify the performance.