Intelligent Recognition Of Species And Living Or Dead’s Microalgae Based On Object Detection

In order to control the species invasion caused by ship ballast water,the International Ballast Water Convention requires the rapid detection of ballast water to ensure that the microorganisms in the treated ballast water meet the requirements of the Convention.At present,there are many methods for the detection of microalgae in ballast water,but they all have some limitations,such as large size of detection equipment,complex steps,high cost and so on.In order to solve the above problems,the microalgae classification method based on deep learning algorithm provides technical support for fast and accurate identification of microalgae species and activity.This paper proposes to study the species identification and activity detection of microalgae based on YOLOv5,which is of great significance to reduce the detection cost and time of ship ballast water.The specific research contents of this paper are as follows:The main results are as follows:（1）In order to improve the accuracy of the algorithm for microalgae species recognition,the effects of image marker ratio and magnification of microalgae image on the model detection results are studied.The experimental results show that:1)compared with the 50%labeled proportional data set,the average accuracy(mAP₅₀)of the verification set is improved by 23%;2)the mAP₅₀ of the model trained by the mixed magnification dataset is 1.8%higher than that of the single magnification on the corresponding verification set.This shows that the detection results of the trained model can be improved by using data sets with high label ratio and mixed magnification.（2）In order to meet the requirements of determining the species of active microalgae,the detection effects of YOLOv5 and the improved YOLOv5 training model on verification sets and different types of test sets were studied.The training model detection results under the two algorithm structures are compared.The experimental results show that:1)the mAP₅₀ of the original model and the improved model for the microalgae verification set are 99.5%and 98.5%respectively;2)the two models can accurately identify the species of active microalgae with few targets and single species of active microalgae with multiple targets;3)for the mixed active microalgae test set,the average accuracy of the original model and the improved model is 97.8%and 94.7%respectively,and the average recall rate is 97.1%and 95.8%respectively.The mAP₅₀ of the two models is 95.2%and 94.6%,respectively.The test results of both models are more than 94%;4)although the mAP₅₀ of the trained model after the improved algorithm is1.1%lower than that of the original model,the number of parameters is reduced from 7.01×10⁶ to 1.39×10⁶,the amount of calculation is reduced from 15.8×10⁹ floating point to 2.5×10⁹floating point,and the detection speed of the model is improved by 150F/s on FPS.Therefore,the microalgae species recognition model based on YOLOv5 and improved YOLOv5 training can distinguish between single species of active microalgae and mixed species of active microalgae,and the detection speed of the improved model is faster,the number of trained parameters and the amount of calculation are lower.（3）In order to meet the requirements of microalgae activity detection,the effect of the improved YOLOv5 training model on the detection of dead and alive microalgae was studied and compared with the artificial method.The results show that:1)the classification accuracy of the model for verification set of dead microalgae and survival microalgae is 83.7%,and the average accuracy is 91.5%;2)in the detection of 100 mixed dead and living microalgae,the average recognition accuracy of the model is 92%,of which the recognition accuracy of dead algae is91.6%and that of living algae is 92.3%.Therefore,the model trained based on the improved algorithm can detect the activity of microalgae.This study can not only identify the types of microalgae in ship ballast water accurately and quickly,but also reduce the detection cost of microalgae and the equipment requirements of model training based on the improved algorithm.it provides an idea for the on-site detection of microalgae in a convenient device.At the same time,the activity detection of microalgae in ship ballast water is studied,and a new detection technology of ship ballast water activity is provided.