Design Of Intelligent Tool Management System

Tool management is related to the quality of aviation maintenance and even the safety of flight and personnel in aviation maintenance.Improving the level of intelligence in tool management is of great significance to the development of aviation.As a representative of the most advanced intelligent methods,this paper studies the tool management system on the basis of deep learning,and applies three image recognition technologies of face recognition,tool recognition and character recognition to the tool management system,then designs and develops a tool management interface by QT.The following is a summary of the main work in this paper:1.Introduced the basic idea of deep learning,and the typical representative convolutional neural network CNN,which has great advantages in the field of image recognition with its unique structure.At the same time,the requirements of the intelligent tool management system are explained.2.In the existing deep learning face recognition algorithm,the simplified version of VGG is selected,that is Lite VGG,and the face recognition verification algorithm is implemented by combining the MTCNN face detection alignment algorithm,then train the original VGG and Lite VGG networks through the training data set.After that,the ten-fold cross-validation method is used to compare and test them on the LFW dataset.By comparing the generated ROC curve and 6000 pairs of face test results,the Lite VGG face recognition method is more effective,more accurate and more suitable for application of the tool management system in this paper.3.Design the maintenance tool recognition algorithm based on Faster RCNN.Create five types of tool datasets including tools with large scale spans by self-built,then implement that by using pre-trained model.The test set is tested,and its m AP can reach 98%,which can basically recognize all tools.For the borrowing and returning of the boxing tool,the image background difference method is used to check whether there is a tool missing,and which includes image processing such as grayscale,morphological corrosion,grayscale reconstruction,and binarization.4.The typical Le Net network is used for tool character recognition.before that,two parts are processed,including character area detection and character segmentation.Character area detection includes edge detection,morphological expansion corrosion,character area screening and so on.Character segmentation consists of pre-segment character preprocessing and character segmentation.Preprocessing needs skew correction,binarization and removing invalid area of characters for making character segmentation more accurate.5.Combining application requirements for intelligent tool management system and background,design database for tool management,including personnel information,tool information and borrowing record table,and use tool management interface which is designed by QT to integrate face recognition,tool recognition and character recognition.The interface program completes the detection experiment,including the face recognition rate,the missed detection rate of the tool and the accuracy of the character recognition.Finally,this paper analyzes and summarizes its shortcomings based on the test results.