A fast identification method for defective shrimp based on deep convolutional neural network

Abstract

The invention discloses a method for quickly identifying defective shrimp based on a deep convolutional neural network, comprising the following steps: 1) improving the first convolutional layer and the second convolutional layer of the traditional LeNet-5, and expanding them into three Layer-parallel network; 2) Add the idea of ​​combination classifier, improve the traditional LeNet-5 classification mode, and form a classification combination layer; 3) Optimize and adjust the network structure with the least parameters and the best recognition rate indicators. The present invention can be directly used in the sample pretreatment link of the prawn breeding and processing plant to remove defective shrimps and realize the online evaluation and measurement of the prawn quality so as to meet the selection and grading requirements of the prawn products.

Description

technical field [0001] The invention relates to the technical field of non-destructive testing of agricultural products, in particular to a method for quickly identifying defective shrimp based on a deep convolutional neural network. Background technique [0002] In the shrimp farming industry chain, impurities or defective shrimps in shrimp products are often removed by manual selection, which is inefficient and the workers are prone to fatigue; at present, labor is generally expensive, and defective shrimps are manually selected It is no longer applicable, and there is an urgent need for efficient and intelligent methods to replace manual automatic sorting of prawns. The core part of automatic prawn sorting is algorithm design and innovation. The quality of the algorithm is directly related to the quality evaluation of shrimp products. Generally, there are defective shrimp and fish and shrimp appendages in the shrimp fished from the pond of the breeding factory, which main...

AI Technical Summary

Problems solved by technology

In the above-mentioned published patents, since different researchers have different cognitions on the same thing, they will more or less incorporate personal cognition factors when designing the shrimp classification algorithm, which will have a negative impact on the algorithm construction. Larger deviation; moreover, highly integrated optimization algorithms need to spend a lot of time and effort to optimize, time-consuming and labor-intensive

[0005] (2) Although some scholars have proposed a variety of shrimp defect identification algorithms in the early stage, the false positive rate has always been high, that is, the probability of misjudging defective shrimp as normal shrimp, especially those with diseased shrimp or necrotic shrimp. If it flows into the normal shrimp population, it will cause serious pollution to the normal shrimp population, resulting in large economic losses and catastrophic consequences

Images