A Low-Cost Detection Method for Nitrite Content in a Mariculture Water Environment Based on an Improved Residual Network

Abstract

Nitrite content is one of the key indicators for measuring the quality of mariculture water and has a crucial impact on the benefits of aquaculture. Most of China’s fisheries are small-scale domestic aquaculture. For economic reasons, farmers generally use chemical colorimetry or rely on life experience (such as whether the water bodies have become turbid or whether aquatic organisms have abnormal or died) to determine the nitrite content in water; however, both methods can easily lead to misjudgment and cause losses. Another more accurate method is spectrophotometry, but the spectrophotometer used is more expensive. This article aims to propose a low-cost and high-precision nitrite detection method. The new method we propose is to first perform a color development reaction using chemical detection reagents, and then use an improved residual network instead of human eyes to determine the nitrite concentration in the water sample. The advantages of this method are the fast response of the chemical reagents and the high accuracy of the machine vision recognition. Our network can achieve an accuracy of 98.3% on the test set. The experimental results indicate that this method can be applied to practical mariculture.