Research On Nail Plate Product Recognition And Positioning Method Based On Deep Learnin

The identification and location method of nail art product is the key technology of automatic sorting system of nail art product.Due to the diversity of size and color of nail art product,the uncertainty of their posture change,as well as the influence of background environment and light,the identification and location methods of nail art product are demanding.Therefore,this paper combines deep learning method to study the identification and positioning method of nail art product.A lightweight detection network FCOS-Mini based on improved Fully Convolutional One-Stage(FCOS)is proposed to realize the identification and detection of nail art product more efficiently.In this method,Ghost Net combined with Effective Channel Attention(ECA)module is used as the backbone network for feature extraction,and Bidirectional Feature Pyramid Network(Bi FPN)module is used to obtain richer multi-scale feature information to improve detection accuracy.The classification category and location information are output by lightweight detection head,and this method is optimized by Generalized Focal Loss(GFL).The experimental results show that FCOS-Mini has fewer parameters and higher detection speed.A lightweight detection network YOLO-Lite based on improved You Only Look Once(YOLO)network is proposed to further improve the identification and detection effect of nail art product.In this method,Shuffle Netv2 with a convolution kernel size of 5 is used as the backbone network to enhance the feature extraction capability,and the spatial pyramid pooling module is combined with cross stage partial network to reduce the number of parameters,and the improved Convolutional Block Attention Module(CBAM)module is added into the multi-scale feature fusion module to improve positioning accuracy.The experimental results show that YOLO-Lite has higher recognition accuracy and detection accuracy.The designed detection network is used to identify the nail art product.The identification results are cut to obtain the nail art product image,which is de-noised and grayed.The contour information is extracted by using Ostu method and Canny edge detection algorithm.Combining the shape characteristics of the nail art product with the least-square method fitting,the position of the grabbing point and the inclination angle of nail art product are determined.The nail art product identification and positioning system based on deep learning is completed,and the interactive interface of this system is designed.The detection algorithm of nail art product and the method to determine the grasping point and tilt Angle are applied to the system for experiments,and the effectiveness of the proposed algorithm is verified.The detection algorithm of nail art product and the positioning method of grasping point and inclination angle are applied to the system and the experiments are carried out to verify the effectiveness of the proposed algorithm.