The Object Detection Algorithm Application Based On Deep-Learning

With the development of computer vision,object detection as an important part of computer vision which also makes a great breakthrough.Traditional object detection method based on sliding windows with no specific region proposal,and the characteristics which need manual design.For complex background,these characteristics don't have a good robustness,which affect the performance of object detection.In recent years,the development of deep learning as a powerful tool can help people achieve better research and exploration in the field of object detection.Because of the powerful feature expression ability of convolutional neural network,it has strong robustness in extracting features,and more and more is applied to object detection algorithm.There are two kinds of object detection methods based on deep learning,one is based on candidate region method and another is based on regression method.The representative methods of these two methods are Faster R-CNN algorithm and SSD algorithm.Compared with the traditional object detection algorithm,the object detection algorithm based on deep learning has a great improvement in speed and precision.In this paper,these two algorithms are studied using on the CAPTCHA data set and the insulator data set.For the CAPTCHA data set,the CAPTCHA recognition framework based on the Faster R-CNN algorithm is proposed and it also confirmed by experiment.Based on the traditional network structure,this paper also improves the accuracy of CAPTCHA recognition by adding network layer number and designing new anchor scheme,and verifies it through experiments.For insulator data set,by comparing between the Faster R-CNN algorithm and SSD algorithm on insulator recognition,the SSD algorithm was selected as an algorithm for insulator detection.At the same time,RANSAC algorithm was used to design the missing insulator detection algorithm.The missing insulator in the image was realized by using the SSD algorithm and the missing insulator detection algorithm.Finally,a missing insulator detection software is written with the programming language,which can realize the missing insulator detection form images,folders and video by automatic loading detection model.