ARM-based fruit type and color sorting method and system

Abstract

The invention discloses an ARM-based fruit type and color sorting method and system. the ARM-based fruit type and color sorting method comprises the following steps that S1, image capture is conductedon various fruits through a camera module in embedded equipment of an ARM architecture; S2, all images are classified once according to the types of the fruits, and then the same type of fruits are classified for the second time according to colors of the fruits; S3, a deep learning framework is built on a PC, the type images and the color images are trained correspondingly through the PC, and atype recognition model and a color classification model are obtained; S4, the fruits are sorted through the embedded equipment of the ARM architecture, image capture is conducted through the camera module, specific to captured images, the fruits are located the fruit types are recognized through a full convolutional network, then color classification is conducted on each fruit through the full convolutional network, and check results are obtained; and S5, according to the check results, a manipulator is controlled automatically to sort the fruits.

Description

technical field [0001] The invention relates to the technical field of image processing and recognition, in particular to an ARM-based sorting method and system for fruit category and color. Background technique [0002] For a long time, the post-harvest processing of fruits is still in the stage of manual sorting, and a lot of manpower is used to manually sort the types and colors of fruits. Manual sorting not only consumes a lot of manpower and time, but also is subject to subjective influences (human visual deviation, subjective cognition, etc.), and sorting errors may occur. [0003] With the development of machine learning technology, machine learning has also begun to be used in the field of fruit sorting, but machine learning has the following shortcomings: 1. It is necessary to pre-set target features, and different features will have a great contrast in the recognition results. Therefore, developers need to have prior knowledge of relevant features; 2. The recognit...

AI Technical Summary

Problems solved by technology

Manual sorting not only consumes a lot of manpower and time, but also is subject to subjective influence (human visual deviation, subjective cognition, etc.), and sorting errors will occur

[0003] With the development of machine learning technology, machine learning has also begun to be used in the field of fruit sorting, but machine learning has the following disadvantages: 1. It is necessary to pre-set target features, and different features will have a great contrast in the recognition results

Therefore, developers need to have prior knowledge of relevant features; 2. The recognition rate of machine learning is not high, and the recognition rate will be lower when the image has color difference, occlusion, etc.

Images