Recognition Of Circuit Diagram In Physics Of Secondary School

The analysis of circuit diagram is a significance part to solve circuit problems for sec-ondary school students and it is imperative to develop a software as an auxiliary tool for circuit teaching.However,it is a challenge for machine to recognize the circuit diagram promptly.At present,there are still two obvious problems about the recognition of circuit diagrams in machine solutions:on the one hand,a lot of time has been spent on the recognition of circuit component symbols,and the circuit diagrams can not be analyzed in time.On the other hand,the accuracy of circuit component symbols recognition is relatively low,and there is a lack of corresponding measures to improve the accuracy.Recognition of circuit diagram in physics of secondary school can be devided into three part,circuit labels recognition,branch lines detection and components recognition.There are three main innovations in this work:1.The tesseract OCR engine is adopted to recognize circuit labels.Hough transform is used to detect the branch lines.2.A novel approach,Faster R-CNN,is applied to recognize the circuit component symbols.Those circuit diagrams are selected from physics books and exercises of secondary school as data sets.The challenges lies in training model to recognize the circuit component symbols and improve improve the accurary of circuit component symbols.3.1n order to solve the problem of multi-objective overlapping miss detection,the penalty function is introduced to reduce its confidence to improve the non-maximum suppression.The calculation methods include linear weighting,Gaussian weighting.The experiment show that the tesseract OCR engine can recognize circuit labels and Hough transform can be used to detect the branch lines.Most importantly,Faster R-CNN can be adapted to recognize more complex circuit component symbols.Besides,the experiments results show that the Gaussian weighting method is the best method to improve the recall rate and solve the problem of multi-objective missed detection.