Fast Self-Adaptive Digital Camouflage Design Method Based on Deep Learning

Abstract

Traditional digital camouflage is mainly designed for a single background and state. Its camouflage performance is appealing in the specified time and place, but with the change of place, season, and time, its camouflage performance is greatly weakened. Therefore, camouflage technology, which can change with the environment in real-time, is the inevitable development direction of the military camouflage field in the future. In this paper, a fast-self-adaptive digital camouflage design method based on deep learning is proposed for the new generation of adaptive optical camouflage. Firstly, we trained a YOLOv3 model that could identify four typical military targets with mean average precision (mAP) of 91.55%. Secondly, a pre-trained deepfillv1 model was used to design the preliminary camouflage texture. Finally, the preliminary camouflage texture was standardized by the k-means algorithm. The experimental results show that the camouflage pattern designed by our proposed method is consistent with the background in texture and semantics, and has excellent camouflage performance in optical camouflage. Meanwhile, the whole pattern generation process takes a short time, less than 0.4 s, which meets the camouflage design requirements of the near-real-time camouflage in the future.

Troop camouflage detection based on deep action learning

“…Troop camouflage in military operations is indispensable to trick and move as close as possible to the opponent, while from the opponent's side constantly trying to extract field conditions from possible enemy camouflage. Separately, the development of artificial intelligence has provided many benefits for recognition and detection purposes [1]. However, the problem of camouflage detection is considered to be difficult to overcome because distinguishing between objects and the same background requires a different strategy [2].…”

“…Unfortunately, the testing of this work uses artificial targets. It is similar with using deep learning; you only look once (YOLOv3) with an average accuracy of 91.55% [1]. Furthermore, it turns out that Xiao et al [1] used a camouflage dataset that was not considered as a vague, for example, a fighter plane with a sky background or a frigate with an ocean background.…”

“…It is similar with using deep learning; you only look once (YOLOv3) with an average accuracy of 91.55% [1]. Furthermore, it turns out that Xiao et al [1] used a camouflage dataset that was not considered as a vague, for example, a fighter plane with a sky background or a frigate with an ocean background. However, the interesting point lies within how to camouflage the detected object according to the background, even though it is not necessary for the attacking troops in battle.…”

Hybrid camouflage pattern generation using neural style transfer method

“…They trained a YOLOv3 model to identify desired targets in the scenes and then, a pretrained deepfillv1 model was utilized to generate primary camouflage texture. Finally, they standardized the primary camouflage texture using the K‐means algorithm 12 …”

“…Finally, they standardized the primary camouflage texture using the K-means algorithm. 12 As the above literature review indicated, all methods generate a CP from scratch based on background images. However, generating a pattern from scratch using a background image tends to lead to high dimensional space of parameters that is time-consuming.…”