Fish key point phenotype data measurement method

Abstract

The invention provides a fish key point phenotype data measurement method, and relates to the technical field of identification analysis. The method comprises the following steps: firstly, shooting side view images and top view images of a plurality of fishes, and obtaining scale scale parameters of the side view images and the top view images; marking key points of the top view image and the side view image by using marking software; training an AIOI-CNet network model through the top view and side view images marked with the key points, and detecting a plurality of customized key points under the top view and a plurality of customized key points under the side view by using the trained AIOI-CNet network model to obtain coordinates of each key point; and according to the key point coordinates on the side view and the top view detected by the AIOI-CNet network model, obtaining phenotype data of fish bodies on the side view and the top view by combining the camera parameter proportions of the side view and the top view. According to the method, the problems of spatial information loss, long training time, low robustness and the like of key point detection by an existing model are solved, and the overall key point detection precision is improved.

Description

technical field [0001] The invention relates to the technical field of identification analysis, in particular to a method for measuring phenotype data of key points of fish. Background technique [0002] In recent years, my country's annual aquatic product aquaculture production has maintained a relatively stable level, and it is the only major fishery country in the world where the total amount of aquaculture products exceeds the total amount of fishing. The output of aquatic products comes from fishing on the one hand, and aquaculture on the other hand. Among them, China's aquaculture industry has a long history, which can be traced back to the Yin Dynasty more than 3,000 years ago. In the process of artificial breeding, a series of "phenotypic data" of aquatic organisms are the main information for farmers to conduct related research on feeding, breeding, and differential grading. Based on the "phenotypic data" of individual fish, researchers can obtain the growth and d...

AI Technical Summary

Problems solved by technology

The network model does not pay attention to the spatial information of the image, does not pay enough attention to the local details of the image, and the overall accuracy is not high

[0017] When the traditional deep neural network model performs key point detection training on objects, the network model does not pay attention to the correlation between key points, such as the relative position in space, so there are false detections and pixel classification errors when detecting key points of objects. Happening

Images