Research On The Feature Recognition Method Of Fabric Pieces

In the automatic sewing process of garment products,the identification of garment fabric pieces is a very important process.Among them,the identification and classification of garment fabric pieces requires a large amount of human and material resources,which hinders the quality of work and the efficiency of classification.Therefore,realizing the automatic identification and classification of garment fabric pieces is one of the quick ways to improve the production efficiency and profit of enterprises.With the widespread development of computer vision applications,the recognition of garment fabric pieces through digital image recognition has become a new way to explore.In this paper,the characteristics of garment fabric pieces are identified and detected,combined with the current deep learning network to identify the fabric,color,and shape of garment fabric pieces.Firstly,select the appropriate camera and light source according to the accuracy required by the experiment.Using Zhang Zhengyou's checkerboard calibration method to shoot the checkerboard at different angles.By calculating the pixel coordinates of the checkerboard corner points,the camera is finally calibrated.Secondly,this paper completes the identification method of the principal components of garment fabric cuts.Through the camera with a magnification of 100 times,the clothing piece information of 6 different fabrics is collected and preprocessed.The VGG16 network,Inception v3 network,and Mobile Net v3 network are selected for training.The training results show that the three networks can achieve the cutting of clothing fabrics.The main component of the film has a classification accuracy of up to 97%.Then identify and detect the shape and color of garment fabric pieces.By self-constructing the data set of the shape and color of garment fabric and calibrating the data set.After the preprocessing method,the three networks of Fastet R-CNN,YOLO v3 and Retina Net were selected to train the apparel fabric pieces in 7 colors and 5 shapes.The experimental results show that the three networks can recognize the apparel fabric pieces well.With positioning effects.Through comparison and analysis of network results,comprehensive recognition accuracy and processing speed,YOLO v3 network was finally selected as the recognition network in this paper.Finally,an automatic fixing and handling device for garment fabric pieces was built and a software for identifying features of garment fabric pieces was designed.Apparel fabric pieces are cooled by a cooling device to achieve a fixed effect by spraying aerosol-applied fabric pieces.After the garment pieces are transported to the workbench for the next identification and production process,the garment device pieces are finally solved by the heating device solid.Through the experimental conveying device,the automatic fixing and conveying of garment fabric pieces can be realized.