딥러닝 기반 유도무기 표적 검출에 적합한 지역적/전역적 적외선 이미지 전처리 기법에 관한 연구
Research on Local and Global Infrared Image Pre-Processing Methods for Deep Learning Based Guided Weapon Target Detection
0
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
본 논문에서는 적외선 이미지에서 딥러닝 물체 탐지를 사용하여 유도무기의 표적 탐지 정확도 향상방법을 연구한다. 적외선 이미지의 특성은 시간, 온도 등의 요인에 의해 영향을 받기 때문에 모델을학습할 때 다양한 환경에서 표적 객체의 특징을 일관되게 표현하는 것이 중요하다. 이러한 문제를해결하는 간단한 방법은 적절한 전처리 기술을 통해 적외선 이미지 내 표적 객체의 특징을 강조하고노이즈를 줄이는 것이다. 그러나, 기존 연구에서는 적외선 영상 기반 딥러닝 모델 학습에서 전처리기법에 관한 충분한 논의가 이루어지지 못했다. 이에, 본 논문에서는 표적 객체 검출을 위한 적외선이미지 기반 훈련에 대한 이미지 전처리 기술의 영향을 조사하는 것을 목표로 한다. 이를 위해 영상과이미지의 전역(global) 또는 지역(local) 정보를 활용한 적외선 영상에 대한 전처리인 Min-max normalization, Z-score normalization, Histogram equalization, CLAHE (Contrast Limited Adaptive Histogram Equalization)에 대한 결과를 분석한다. 또한, 각 전처리 기법으로 변환된 이미지들이 객체 검출기 훈련에미치는 영향을 확인하기 위해 다양한 전처리 방법으로 처리된 이미지에 대해 YOLOX 표적 검출기를학습하고, 이에 대한 분석을 진행한다. 실험과 분석을 통해 전처리 기법들이 객체 검출기 정확도에영향을 미친다는 사실을 알게 되었다. 특히, 전처리 기법 중에서도 CLAHE 기법을 사용해 실험을 진행한결과가 81.9%의 mAP (mean average precision)을 기록하며 가장 높은 검출 정확도를 보임을 확인하였다.
Abstract
In this paper, we explore the enhancement of target detection accuracy in the guided weapon using deep learning object detection on infrared (IR) images. Due to the characteristics of IR images being influenced by factors such as time and temperature, it's crucial to ensure a consistent representation of object features in various environments when training the model. A simple way to address this is by emphasizing the features of target objects and reducing noise within the infrared images through appropriate pre-processing techniques. However, in previous studies, there has not been sufficient discussion on pre-processing methods in learning deep learning models based on infrared images. In this paper, we aim to investigate the impact of image pre-processing techniques on infrared image-based training for object detection. To achieve this, we analyze the pre-processing results on infrared images that utilized global or local information from the video and the image. In addition, in order to confirm the impact of images converted by each pre-processing technique on object detector training, we learn the YOLOX target detector for images processed by various pre-processing methods and analyze them. In particular, the results of the experiments using the CLAHE (Contrast Limited Adaptive Histogram Equalization) shows the highest detection accuracy with a mean average precision (mAP) of 81.9%.
참고문헌 (24)
- Tsung-Yi Lin, "Focal Loss for Dense Object Detection", Proceedings of the IEEE International Conference on Computer Vision, pp.2980~2988 (2017)
- "YOLOX : Exceeding YOLO Series in 2021" (2021)[URL]
- Hao Zhang, "DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection", International Conference on Learning Representations, pp.1~23 (2022)
- Janghoon Kim, "A Study on Sustainable Competitive Advantage of Defense Venture Companies: Focusing on the Case of i3system Inc", Seminar and Discussion of the Korea Trade Association, pp.213~229 (2017), Chungnam University
- Akula, Aparna, "Thermal imaging and its application in defence systems", AIP conference proceedings, pp.333~335 (2012)
- Zixuan Zhang, "YOLO-IR-Free : An Improved Algorithm for Real-Time Detection of Vehicles in Infrared Images", Sensors, 23, pp.1~23 (2022)[DOI]
- Du, Shuangjiang, "FA-YOLO: An improved YOLO model for infrared occlusion object detection under confusing background", Wireless Communications and Mobile Computing, 2021, pp.1~10 (2021)[DOI]
- Lixin Zhang, "Real-time segmentation method of billet infrared image based on multi-scale feature fusion", Scientific Reports, 18, pp.1~7 (2022)[DOI]
- "A Multi-scale Information Integration Framework for Infrared and Visible Image Fusion" (2023)[URL]
- Zhengzheng Tu, "RGBT Salient Object Detection : A Large-Scale Dataset and Benchmark", IEEE Transactions on Multimedia, 25, pp.4163~4176 (2022)[DOI]
- Soonmin Hwang, "Multispectral Pedestrian Detection : Benchmark Dataset and Baseline", Proceedings of the Conference on Computer Vision and Pattern Recognition, pp.1037~1045 (2015)
- Chenglong Li, "RGB-T object tracking : Benchmark and baseline", Pattern Recognition, 96, pp.1~12 (2019)[DOI]
- Chenglong Li, "LasHeR : A Large-Scale High-Diversity Benchmark for RGBT Tracking", IEEE Transactions on Image Processing, 31, pp.392~404 (2021)[DOI]
- R. Gonzalez, "Digital Image Processing" (2009), Pearson Education
- S. M. Pizer, "Contrast-limited adaptive histogram equalization : speed and effectiveness", Proceedings of the First Conference on Visualization in Biomedical Computing, pp.337~345 (1990)
- Huang, Lei, "Normalization techniques in training dnns:Methodology, analysis and application", IEEE transactions on pattern analysis and machine intelligence, 45, pp.10173~10196 (2023)[DOI]
- Kandanaarachchi, Sevvandi, "On normalization and algorithm selection for unsupervised outlier detection", Data Mining and Knowledge Discovery, 34, pp.309~354 (2019)[DOI]
- Jae-Won Lee, "Histogram Equalization based on Differential Compression for Image Contrast Enhancement", JOURNAL OF BROADCAST ENGINEERING, 19, pp.96~108 (2014)[DOI]
- Byong-Seok Min, "A Novel Method of Determining Parameters for Contrast Limited Adaptive Histogram Equalization", Journal of the Korea Academia-Industrial cooperation Society, 14, pp.1378~1387 (2013)[DOI]
- Hou, Fujin, "Review on infrared imaging technology", Sustainability, 14, pp.1~26 (2022)[DOI]
- Shu, Hai, "Sensitivity analysis of deep neural networks", Proceedings of the AAAI Conference on Artificial Intelligence, pp.4943~4950 (2019)
- Hyunji Cho, "EThe Clip Limit Decision of Contrast Limited Adaptive HistogramEqualization for X-ray Images using Fuzzy", Journal of Korea Multimedia Society, 18(7), pp.806~817 (2015)[DOI]
- Joseph Redmon, "You Only Look Once : Unified, Real-Time Object Detection", Proceedings of the Conference on Computer Vision and Pattern Recognition, pp.779~788 (2016)