Effective image processing-based technique for frost detection and quantification in domestic refrigerators

Hammad ur Rahman,Hassan Akbar,Anjum Naeem Malik,Tahir Nawaz,Ismail Lazoglu
DOI: https://doi.org/10.1016/j.ijrefrig.2024.01.026
IF: 4.14
2024-01-28
International Journal of Refrigeration
Abstract:Frost accumulation is a common problem when moisture in the air condenses and freezes on surfaces like heat exchange tubes of refrigeration units. Frost accumulation negatively impacts heat exchange by disrupting the process, reducing system efficiency, and causing operational issues. Therefore, defrosting is mandatory to maintain the rated performance; however, modern automatic defrosting systems rely on sophisticated sensors for frost quantification. These sensors are susceptible to degraded performance with the passage of time under varying environmental conditions. To this end, we introduce a robust and generic image processing-based solution that relies on building a data-driven regression-based model for frost detection and thickness estimation. We evaluated the effectiveness of the proposed method on a newly collected dataset with encouraging performance in terms of a low error margin of 13.69% when compared to conventional capacitive and photoelectric sensors-based frost thickness estimation with error margins of 15.17% and 17.5%, respectively. Similarly, other image processing-based methods, such as Global thresholding, Adaptive mean, and Adaptive gaussian thresholding for segmentation, were compared with the proposed method. Deviations in the error margins were found to be 19.94%, 28.96%, and 27.85%, respectively. These findings highlight the appropriateness of employing K-means for estimating frost thickness.
engineering, mechanical,thermodynamics
What problem does this paper attempt to address?