Comparison of High Adaptability and Low-Cost Intelligent Recognition Method for Evaporator Frosting Status Based on Image Texture Features

Unnecessary or delayed defrosting results in increased energy consumption, reduced stability, and increased failure rates in refrigeration and heat pump units. Accurately identifying the frost status and timely defrosting are important for improving the performance of refrigeration and heat pumps. Frost status identification methods based on digital and intelligent technologies have shown significant potential. However, existing technologies have significantly reduced accuracy in complex real-world conditions and require urgent improvement. In this paper, we proposed an intelligent recognition method based on the texture features of evaporator surface images. We used a gray-level co-occurrence matrix to extract texture features and combine them with the extreme learning machine optimized by the sparrow algorithm for classification. This is expected to mitigate the impact of external conditions, such as shooting angles and light intensity, thereby achieving strong adaptability. An experimental setup was established to collect 4125 images of the evaporator in three different frost states under complex working conditions, and the proposed method was validated and compared. The results showed that the accuracy of the method in identifying different conditions can reach 95%, which is significantly higher than that of existing methods by 5-35%. Furthermore, this method has high stability and low cost thereby demonstrating great potential for practical applications.