Nowadays, visual recognition has been applied in various fields, with most factories and many business enterprises constantly introducing artificial intelligence. Since the accuracy of visual recognition technology for assessing the pass rate has been found to be undesirable. In this paper, LST-CBIR detection algorithm for prefabricated pizzas is proposed based on content-based image retrieval (CBIR) system with weights trained by the least squares method. And three feature extraction method of pizza images is described. By extracting the color, shape, and texture features of pizza images and combining them into a comprehensive feature set, the quality of prefabricated pizzas is determined. The weights of the three features are calculated using the least squares method. A pizza quality recognition model based on CBIR is established to identify the quality of prefabricated pizza by calculating similarity between standard sample and tested sample. This approach effectively improves the efficiency and accuracy of pizza quality determination. Experimental results are shown that the LST-CBIR pizza quality recognition method proposed is feasible and superior, providing a new solution for the quality inspection of prefabricated pizzas and applications in machine vision.