Detection and State Classification of Bolts Based on Faster R-CNN

摘要

With the increase of the training frequency of aircraft, the frequency of daily maintenance work also increases dramatically. As key connectors in aircraft, the demand for bolts also significantly increases. Hardness is an important performance index of bolts. At present, the manual inspection method cannot meet the requirement for the hardness test of the large number of the bolts. Therefore, there is an urgent need to develop an automated hardness detection system. In this paper, a bolt state detection method based on Faster R-CNN is proposed, which is an important part of an automated hardness detection system and can assist the robot grasping the bolt and placing it on the hardness tester. This method can detect bolts on the pallet and classify their states. The mean average precision (mAP) of this model used in the bolt dataset is 96.44%.