Design And Implementation Of Image-based Marine Microalgae Identification System

Marine microalgae is a kind of primitive Marine phytoplankton,which is simply called Marine microalgae.It is not only the main Marine living resources in the Marine ecosystem,but also an important primary producer.Many kinds of Marine microalgae have considerable economic value.In the process of studying Marine microalgae,the first thing is to detect the Marine microalgae.In this paper,the digital image processing technology is used to automatically identify and classify Marine microalgae.Firstly,the classification and identification of Marine microalgae by traditional identification methods are studied,and the image binarization is improved by adopting the adaptive threshold binarization method.In the detection part,Hu moment feature and cosine similarity are fused to avoid the influence of size difference of target object on detection.Secondly,based on the research of the deep learning algorithms,found that the algorithm in the application of Marine micro algae recognition than traditional algorithm has high precision,and to this end,this article will study the migration method with yolov3 algorithms,through the experiment proves that the Marine micro algae under the circumstances of less training data,this method has good effect on Marine micro algae recognition.At the same time,in order to further improve the recognition rate and reduce the amount of calculation,this paper,based on the characteristics of small calculation amount of the deep separable convolution algorithm,gives the Gaussian Yolov3 algorithm based on the deep separable convolution.The improved mAP improves by 1.45%,and the FPS reaches 42.35,increasing by 19.95.Finally,this paper designed a set of Marine microalgae classification and identification system,and built a set of user management module,user operation module,background database storage module.Through experiments,it verified that the system designed in this paper has certain practicability.