基于YOLO v3的生猪个体识别方法

为实现高效的猪只个体识别,提出一种基于机器视觉的生猪个体识别方法。通过采集母猪和仔猪个体图像,对图像进行扩充和筛选,制作训练集和测试集。试验采用基于YOLO v3的识别模型,并与Faster RCNN和SSD模型识别结果进行比较,结果表明:对仔猪的识别平均精度均值达89.65%,准确率达95.99%,召回率达84.09%。对母猪的识别平均精度均值达95.16%,准确率达96.00%,召回率达96.00%。相较于Faster RCNN,该模型的识别速率是其7倍以上,相较于SSD,该模型的平均精度均值提高9%,说明该模型在识别速率和识别精度上都达到了较高水平。该研究可为猪只个体智能识别、数据监测及养殖信息化等提供理论依据。

Individual pig identification method based on YOLO v3

In order to realize efficient pig individual recognition,a pig recognition method based on machine vision was proposed.Training sets and test sets were made by collecting individual images of sows and piglets,and the images were expanded and screened.The identification model based on YOLO v3 was adopted and compared with the identification results of the Faster RCNN and SSD models.The results showed that the mean average precision of pig lets identification was 89.65%,the accuracy rate was 95.99%,and the recall rate was 84.09%.The mean average precision of sows identification was 95.16%,the accuracy rate was 96.00%,and the recall rate was 96.00%.The recognition rate of the model is 7 times higher than that of the Faster RCNN,and the average accuracy of the model is 9 percentage points higher than that of the SSD.The model reaches a high level in both the recognition rate and the recognition accuracy.This study can provide theoretical basis for intelligent identification,data monitoring and breeding informatization of pigs.