Comparative Research And Application Of Three Target Detection Algorithms In Location Recognition Of Solder Joint Of Automobile Door Plate

With the increasing demand of domestic automobile market,automobile production efficiency is facing a huge challenge.Welding is a very important part of automobile production.Although the welding robot has been used in automobile production line,the traditional welding robot uses the way of artificial teaching to work.The robot itself can't recognize the position and type of welding points of automobile door panel intelligently,which greatly reduces the production efficiency of automobile welding production line.In order to improve the production efficiency,this paper studies the target detection algorithm related to deep learning,and applies it to the position detection of automobile door panel welding joints.The research content of this paper includes the following three parts:First of all,this paper deeply studies the theoretical knowledge of neural network and deep learning.It mainly includes the basis of artificial neural network,the improved learning method of artificial neural network,the commonly used convolutional neural network architecture,the optimization algorithm of deep learning,etc.The common convolutional neural network architecture can improve the performance of target detection,and the deep learning optimization algorithm can be used to improve the training efficiency of neural network.Secondly,this paper studies the application effect of Faster R-CNN algorithm based on candidate region and YOLO algorithm based on regression on the position recognition of automobile door panel solder joints.Using the same data set and experimental environment,it mainly verifies the recognition accuracy and detection time of solder joints to determine whether it can meet the actual production requirements.Finally,aiming at the problem that Faster R-CNN algorithm and YOLO algorithm can't balance the recognition accuracy and detection time of solder joints,this paper studies the recognition effect of SSD algorithm based on the concept of region on the position of solder joints of automobile door panel.Then,for the particularity of the research example in this paper,an improved method for smalltarget detection is proposed.Through the experiment,we can draw the conclusion that the improved SSD algorithm can quickly detect and recognize the target on the basis of maintaining high recognition accuracy,and can meet the requirements of real-time and accuracy in the production process.