Scholarlink

Deep learning (DL) and Machine learning(ML) based object detection and classification have become more popular in many applications, such as medical image processing, face identification, self-driving vehicles, pedestrian detection, and security monitoring systems, in recent years. Traditional existing AI algorithms have limitations like image occlusion and distortion that is caused by interruption of noise signals and also due to smoke, fog and low lighting conditions. This proposed system overcomes these drawbacks by using the thermal cameras that acquires heat radiations from objects. Thermal camera works irrespective of environmental and lighting condition, which widens this research scope. Moreover traditional algorithms are computationally expensive and it utilizes more time and resources. Considering all these facts, this research work proposes a systematic pre-processing module for eradicating the noise signals from thermal image. The performance of the pre-processing module is validated with Structural Similarity Index (SSIM) and Peak signal to Noise Ratio (PSNR). Further, the pre-processed thermal image is classified using the ensemble YOLO models and it is predicted that the YOLOv3 model is suitable for this research work. The YOLOv3 model has a decent accuracy rate of 92% with less computational time (~ 0.5764 s).

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Object Identification and Classification from Thermal Images Using the Ensemble Learning Across Various YOLO Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Mary Amirtha Sagayee G,
  • Nancy V,
  • Rolando Jr Lontok

摘要

Deep learning (DL) and Machine learning(ML) based object detection and classification have become more popular in many applications, such as medical image processing, face identification, self-driving vehicles, pedestrian detection, and security monitoring systems, in recent years. Traditional existing AI algorithms have limitations like image occlusion and distortion that is caused by interruption of noise signals and also due to smoke, fog and low lighting conditions. This proposed system overcomes these drawbacks by using the thermal cameras that acquires heat radiations from objects. Thermal camera works irrespective of environmental and lighting condition, which widens this research scope. Moreover traditional algorithms are computationally expensive and it utilizes more time and resources. Considering all these facts, this research work proposes a systematic pre-processing module for eradicating the noise signals from thermal image. The performance of the pre-processing module is validated with Structural Similarity Index (SSIM) and Peak signal to Noise Ratio (PSNR). Further, the pre-processed thermal image is classified using the ensemble YOLO models and it is predicted that the YOLOv3 model is suitable for this research work. The YOLOv3 model has a decent accuracy rate of 92% with less computational time (~ 0.5764 s).