Video Flame Recognition Based on Feature Fusion and Extreme Learning Machine
| Autor: | Tian Zhijia, Chen Zhong, Ke Wang, Hongjun Ding |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Color histogram
Feature fusion Gravity center Computer science business.industry Flame detection ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Pattern recognition GeneralLiterature_MISCELLANEOUS Roundness (object) Extractor Artificial intelligence business Operating speed ComputingMethodologies_COMPUTERGRAPHICS Extreme learning machine |
| Zdroj: | CACRE |
| DOI: | 10.1109/cacre50138.2020.9230253 |
| Popis: | This paper proposes a video flame detection method based on Extreme Learning Machine (ELM). Visual Background Extractor++(ViBE++) algorithm is used to extract the dynamic foreground features of flame video images, and combined with color histogram threshold analysis, the flame region in the image is segmented. By extracting and processing the flame geometrical features such as roundness, sharpness and gravity center, the dynamic and static features of flame image are fused, thus the suspected flame area is screened out. Using the geometric features of suspected flame area and based on ELM model, the training and classification of sample sets are performed. Experimental results show that this method has higher operating speed and accuracy under the condition of less environmental interference. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|