Research on postharvest tomato freshness recognition method based on RGB-S and ResNet34

The accurate identification of postharvest tomato freshness is critical for fruit growers to plan their postharvest storage, transportation, and wholesale processes. In this study, a method based on improved frequency-tuned (FT) visual saliency detection and ResNet34 model is proposed for nondestructive identification of postharvest tomato freshness. The L*, Y, and H components were extracted as effective features to be introduced into the original FT algorithm by performing color space analysis and image processing operations on tomatoes variation images with different freshness levels. The improved FT algorithm was utilized to obtain visual saliency maps, which were combined with the corresponding RGB image information to form four-dimensional data. The ResNet model was improved as a four-channel model to realize the classification of tomato freshness. The experimental results show that the accuracy, precision, and recall of the method are 98.38%, 98.69%, and 98.32%, respectively. The detection speed of a single image is 0.0326 s. The results of the study demonstrated that the proposed method for recognizing postharvest tomato freshness has effectiveness and real-time performance and can provide technical support to the fruit and vegetable production and processing industries and consumers when shopping for fresh tomatoes..