Research And Implementation Of Steel Pipe Sign Information Recognition Based On Deep Learning

The daily shipments and purchases of steel pipes by steel pipe manufacturers and steel pipe reprocessing manufacturers in my country are extremely large,which brings huge challenges to steel pipe information statistics.At present,the vast majority of steel pipe manufacturers still use the method of manual recording of steel pipe information by workers,which leads to the occurrence of omissions and errors in the statistics of steel pipe information,which in turn affects the production efficiency and correct traceability of steel pipes.With the rapid development of big data and artificial intelligence technology,great progress has been made in the field of computational vision technology,and various algorithm models emerge in an endless stream,but so far almost no algorithm models have been applied to steel pipe information statistics.There are two main carriers of steel pipe information,one is the spray code on the steel pipe body,and the other is the steel pipe that is placed in bundles during storage and transportation.The steel number,furnace number and batch number of each bundle of steel pipes are all It is the same,each bundle of steel pipes carries a sign,and the information of the bundle of steel pipes is recorded on the sign.This article is mainly aimed at detecting,identifying and recording the information of steel pipe signs.The information recognition of steel pipe signs is mainly divided into two aspects: text detection and text recognition.In terms of text detection,this paper first studies the OCR technology based on traditional image algorithms,and uses traditional image processing algorithms to detect and segment the signs.Then the CTPN algorithm based on neural network is used to detect the text of the signs,and then the advantages and disadvantages of the two algorithms are analyzed,and a text detection algorithm specially designed for steel pipe signs,that is,the ORB+CTPN combination algorithm is proposed based on this.Finally,it is verified by experiments that the text detection accuracy of the ORB+CTPN combination algorithm on steel pipe signs is higher than that of the traditional image algorithm and the CTPN algorithm alone.In terms of text recognition,the CRNN algorithm based on neural network is used to recognize the sign text,and the latest dynamic ReLU is used to replace the static ReLU to further improve the accuracy of text recognition.After the text detection and text recognition tasks of steel pipe signs are completed,in order to make the algorithm better landing and application,this paper uses PyQt to design a steel pipe sign recognition system,and deploys the system on an industrial tablet.Take pictures of the sign,and then click the identification button on the interface to identify the text information,which is very convenient for workers to work in the factory in real time.