Automatic fruit identification method based on attention YOLOv5 model

Abstract

The invention discloses an automatic fruit identification method based on an attention YOLOv5 model. The method comprises the following steps: 1, preprocessing an original image of a data set; 2, inputting a backbone network to extract features, and using an SENet attention module to obtain a one-dimensional vector corresponding to a channel as an evaluation score; 3, the evaluation scores act on corresponding channels of the feature map through multiplication operation, and effective features used for fruit recognition are obtained; 4, carrying out the fusion of the features through a Feature Pyramid Network structure and a Path Aggregation Network structure, and carrying out the fusion of the features through the Feature Pyramid Network structure and the Path Aggregation Network structure; and 5, predicting the image features, considering the relationship between the height-width ratio and the central point of the target frame and the prediction frame by using CIOU, improving the prediction precision, and respectively outputting identification results according to large, medium and small targets.

Description

Technical field [0001] The present invention belongs to the field of fruit recognition, in particular to an automatic fruit recognition method based on the attention YOLOv5 model. Background [0002] In daily life, supermarkets and wet markets need to be weighed and priced manually, which not only takes time to operate, the pricing process is too long, but also reduces the purchase efficiency of customers. Therefore, it is very necessary for supermarkets and wet markets to combine fruit weighing, automatically calculate the price of fruits, and improve the efficiency of customer purchases. [0003] At present, the mainstream method mainly has the following problems: 1, the model tends to the global information of the object, and it is easy to ignore some key and important fruit local information; 2, the target box and the prediction box coincide without considering the interrelationship between them, which can easily lead to inaccurate prediction results. Contents of the Invent...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide an automatic fruit recognition method based on the attention YOLOv5 model, so as to solve the problem that the existing model tends to the global information of the object, and it is easy to ignore some key and important local information of the fruit and the coincidence of the existing target frame and the prediction frame The interrelationship between them is not taken into account, which is likely to lead to technical problems of inaccurate prediction results

Images