Study Of Intelligent Feeding Decision Method Based On Near Infrared Machine Vision For Fish In Aquaculture

In aquaculture,the feeding efficiency of fish is of great significance for improving production and reducing costs.In recent years,automatic adjustments of the feeding amount based on the needs of the fish have become a developing trend.The purpose of this study was to achieve automatic feeding decision making based on the appetite of fish.In this study,an intelligent feeding decision method based on near infrared machine vision and neuro-fuzzy model was proposed.This system provides an important contribution to realize the real-time control of fish feeding processes on demand,and it lays a theoretical foundation for developing fine feeding equipment and guiding practice.The paper focuses on the following aspects:Due to the low and uneven illumination that is typical of an aquaculture system,the images collected from fish farm always have low brightness and contrast.A near infrared image preprocessing method based on the reflective frame classification and adaptive contrast enhancement was proposed.Through the design of intelligent and adaptive algorithms,the running speed of the method was improved under the premise of ensuring that the method still has a high classification accuracy.Meanwhile,the improved Multi-Scale Retinex algorithm was used to eliminate the influence of low and non-uniform illumination,and then the gray scale non-linear transformation can be realized by normalized incomplete Beta function.The results showed that the average classification accuracy,FPR and FNR for two types of reflection frames were 96.34%,4.65% and 2.23%,respectively.And the image contrast was increased by more than 3 times,furthurmore,the average recognition rate of fish was improved about 7.9%.Therefore,this method could function as an important step in pre-processing the images in aquaculture.In view of the serious overlap of images captured in aquaculture,which can often affects the subsequent recognition rate,a method of segmentation of fish overlapping images has been proposed.After the steps of overlap region determination,corner extraction and skeleton extraction,the linear equation was used to segment overlapped regions.The results show that the average error rate of this method is 10% and the average segmentation efficiency is 90%.Thus,the proposed method achieves better performance in segmentation accuracy and effectiveness.This method can be applied to multi-target segmentation and fish behavior analysis systems,and it can effectively improve recognition precision.Fish feeding behavior holds important information for the aquaculturist.In this paper,a method for quantifying the feeding behavior and extracting the index for fish based on the near infrared machine vision was proposed.Flocking index of fish feeding behavior(FIFFB)and snatch intensity of fish feeding behavior(SIFFB)were extracted by Delaunay Triangulation and image texture respectively.The results showed that the behavioral changes during fish feeding can be accurately expressed by FIFFB and SIFFB.Compared with the results of artificial expert scoring,the linear correlation coefficient of FIFFB can reach 0.945,and the linear correlation index between SIFFB and area method is 0.876.Which provides an effective way to quantify fish feeding behavior and can be used to guide production practices.In view of the serious waste existing in the current feeding process,an intelligent feeding decision method based on near-infrared machine vision and neuro fuzzy model was proposed.The FIFFB and SIFFB were used as input of adaptive network-based fuzzy inference system(ANFIS),which realizes intelligent decision making during fish feeding process(continue or stop feeding).The results indicated that the feeding decision accuracy of the ANFIS model was 98%.In addition,the feed conversion rate can be improved by10.77% and water pollution can also be reduced.This system provides an important contribution to realizing the real-time control of fish feeding processes on demand,and it lays a theoretical foundation for developing fine feeding equipment and guiding practice.