| Piano soundboard,also known as resonance board,is one of the most important parts of piano.Different types of soundboard have obvious influence on pianistic sound quality.Soundboard classification is an important part in the piano production process.The difference in texture and color naturally formed on the surface of the soundboard is an important basis for classifying soundboards.In order to respond to the user’s choice of piano acoustic quality,piano processing plants need to identify and classify according to the surface features of the soundboard.In this situation,traditional manual inspection can no longer meet the rapid development of the musical instrument manufacturing industry.In response to this phenomenon,this article uses image processing technology to conduct a detailed study on the entire piano soundboard real-time recognition process.The main work is as follows:1)Design of soundboard imaging acquisition system: In order to meet the actual production requirements of the company,determine the soundboard classification standards,design the process flow of the automated testing platform,introduce the corresponding mechanical and electrical hardware structure,and formulate the software development plan based on digital image processing technology through the actual needs of the project.2)Data collection and preprocessing of soundboard image: The soundboard material used in this article comes from a piano processing factory,and the soundboard image is captured by an industrial area array camera.In order to accurately locate the soundboard area,a component threshold segmentation method based on RGB color space is proposed to determine the soundboard boundary position.In order to improve the image quality,firstly perform weighted grayscale processing on the image of the soundboard area,and then use bilateral filters for noise reduction and edge preservation processing,through the contrast limited adaptive histogram equalization algorithm(CLAHE)to further highlight the texture information of the image of the soundboard area,which is conducive to the extraction of texture features in the later stage.3)Extraction of Soundboard texture feature: From the perspective of soundboard texture information,a real-time classification algorithm based on multi-feature fusion is proposed.Extracted multiple texture feature values based on the image surface gray level,co-occurrence matrix(CLCM),and fractal theory.Then constructed an adaptive boosting(Ada Boost)classification model for the soundboard’s detection and identification.4)Extraction Soundboard color feature: In the design of the soundboard color classification algorithm,a feature extraction method based on the color moment of the RGB color model and the component mean value of the HSV color model is designed for the color difference of the soundboard to ensure efficient and accurate classification.A soundboard color classification algorithm based on Random Forest(RF)is proposed to ensure the criteria of efficient.In order to verify the effectiveness and accuracy of the soundboard real-time recognition system,a variety of comparative tests were carried out to make judgments.In the soundboard image enhancement algorithm,comparing different image enhancement algorithms to verify the algorithm in this paper has a better effect on highlighting image texture of the soundboard.In order to verify the accuracy of the soundboard classification algorithm in this article,a comparative experiment based on different feature extraction algorithms was carried out in terms of soundboard texture and color.Experiments show the algorithm used in this paper meets the real-time classification of soundboard texture and color,the average recognition rate of soundboard texture reaches 93.59%,and the average recognition rate of soundboard color reaches 94.18%,reaching the actual production standard of piano enterprises. |
