High resolution object detection algorithm based on parallel

Abstract

With the continuous development and improvement of the aviation information warfare system, target detection technology has also become a key part of the airborne system to perceive the environment. Traditional target detection technology is now difficult to meet the requirements of high precision and high real-time performance in airborne scenarios. With the continuous development of deep learning technology, neural network has become the latest method to deal with object detection task, which greatly improves its accuracy and processing efficiency. However, due to the different targets detected in airborne scenes, the scale of data needed to be processed by using neural networks to process target detection tasks also expands dramatically, and the computing resources provided by single chip are already difficult to meet the needs of target detection algorithm execution in airborne environment. This paper proposes a set of high-precision target detection algorithms based on parallelism, which greatly improves the precision and processing efficiency of the target detection algorithm in airborne scenarios.