Research On The Underwater Biological Information Acquisition System Of Crab Larvaes Based On Computer Vision

With the development of modern aquaculture,the informatization in breeding process has become a demand.In the process of artificial breeding of swimming crab Portunus trituberculatus,the nursery environment and the state of crab larvaes need to be controlled,which can help breeders make adjustments in time to avoid loss.Using computer vision technology to collect biological information,not only reduces the interference to the creatures but also improves the objectivity.It can help breeders reduce risk and increase production.Based on the behavior characteristics of crab larvaes,this paper studied the underwater information acquisition system of crab larvaes by computer vision technology.The details of research work and results are as follows:First,the recognition of crab larvaes in the actual breeding environment is the basis of the practical application of crab information collection.In this study,different recognition methods were adopted for crab larvaes at different growth stages.After modeling the background of the video,foreground images with motion characteristics were obtained.These images contained crab larvaes and various impurities.For the zoaea and small impurities,the contour and brightness features of their images were extracted.And the SVM classifier was used to classify the images to realize the recognition of the zoaea.For the megalopa and the impurities at the same size,Haar-like features in their images were extracted.And the Adaboost classifier was used to classify the images to realize the recognition of the megalopa.The results showed that the recognition accuracy of the zoaea was higher than 91.56%,and the recognition accuracy of the megalopa was higher than 93.21%.The result indicated that targeted crab larvaes recognition method can achieve recognition function.Second,The health status of crabs is extremely important to the breeding process,and the quantification of the health status is helpful to the breeder's judgment.There is a positive correlation between the health status and phototaxis ability of crab larvaes.The phototaxis ability of the crab larvaes was transformed into numerical value,so that the healthy status of crab larvaes was quantified.The number of crab larvaes near the light source was calculated and the curve of aggregation number was plotted.The slope of the curve represented the phototaxis ability of the crab larvaes.In the process of counting the number of crab larvaes,different statistical methods were adopted for the different degree of aggregation.For low-density images,linear regression was used to estimate the number of crab larvaes by the number of foreground pixels.For high-density images,the density levels were divided through the competitive network,and then the range of the number of crab larvaes was estimated by determining the density level of the current image.The experimental results showed that the accuracy of this statistical method was 93.95%.It can be used to quantified the status of crab larvaes.Thirdly,Crab production and breeding profits are closely linked,and the total amount of crab larvaes is important for breeders.An underwater video capture device was designed to collect video in different layers by adjusting the height of the recording device.After the acquisition,the crabs in video were recognized and counted to be the sample data.And then the probability density of the crab in the water was obtained through the nuclear density estimation.Finally,the total value of the crabs was estimated based on the probability density function and the sample value.The experimental results show that the accuracy of the estimation method in the small pond was 82.14%.But the accuracy was reduced in large ponds.Further improvement is expected by adjusting the sampling scheme.This paper designed the collection equipment and operation plan for the status and total amount of crab larvaes in the breeding environment,and presented methods for analyzing and processing the images of crab larvaes.These made underwater biological information acquisition to be intelligent.