Detection of Mobile Phone Screen Defect Based on Faster R-CNN Fusion Model

With the increasing prevalence of mobile phone products, the quality of mobile phones has become more and more vital, especially the quality of mobile phone screens. Therefore, the accurate detection of screen defects is essential. However, traditional manual and machine vision method detection have the problems of low detection accuracy and slow efficiency. Based on the problems of the above detection methods, this paper proposes the detection method based on Faster R-CNN fusion model to improve the detection performance. First, the designed image preprocessing methods, U-Net and ResNet50, are combined and used as the feature extraction part of the whole network with the aim of obtaining more distinct feature information. Secondly, the region proposals the network part to generate initial anchors of specific size by K-means clustering algorithm instead of manual setting, which can have an improvement on the accuracy and rate of classification and position prediction of bounding boxes. The experimental show that on our dataset, improved Faster R-CNN achieves 83% mAp, which proving the applicability of the model proposed in this paper.