基于Faster R-CNN的精密零部件的识别方法

在航空航天领域中，惯性陀螺等精密器件装配精度要求较高，目前大多采用人工装配的方法，装配效率低、装配过程受人主观影响大。针对上述存在的问题，采用基于Faster R-CNN模型的目标识别算法，通过VGG16特征提取网络提取特征信息，在模型训练过程中利用COCO数据集的深度网络模型进行迁移训练，防止模型过拟合并加速参数的训练过程。同时，该方法还与其他深度学习模型以及传统的目标识别算法进行了对比，在自建的数据模型测试集上进行试验。结果表明，基于VGG16的Faster R-CNN目标识别模型在复杂环境及物体发生遮挡的情况下对于惯性陀螺的识别具有明显的优势，准确率可达到87.80%，召回率80.30%，识别速度可达到15FPS，能够满足实时性要求。

Detection Method of Precision Parts Based on Faster R-CNN

In the field of aeronautics and astronautics, the assembly accuracy of precision devices such as inertial gyroscope is required to be high. At present, most of them are assembled manually, which has low assembly efficiency and the assembly process is easily influenced by human subjective. In view of the above problems, this paper adopts the objection detection algorithm based on Faster R-CNN, extracts the feature information through vgg16 network, and uses the depth network model of COCO data set for migration training in the process of model training to prevent the model from over-fitting and accelerating the training process of parameters. At the same time, the method is compared with other deep learning models and traditional algorithms, and is tested on the self built data model test set. The results show that the Fast R-CNN object detection model based on vgg16 has obvious advantages for the detection of inertial gyro in the complex environment and when the object is blocked, the accuracy can reach 87.80%, the recall rate 80.30%, and the recognition speed can reach 15fps, which can meet the real-time requirements.