Research And Application Of Convolutional Neural Network Algorithm Based On Integrated Leamingin Small-scale Scene Classification

The image information contains a lot of information and the content is easy to understand and concise.With the rapid development of society and the rapid advancement of Internet technology,image data has become a very important research content in industry and scientific research.In one of the most popular directions in computer vision,the classification and recognition of images is the foundation and one of the most important parts of computer vision.Image scene classification is an important direction in image classification.The two most famous directions for scene classification are natural scene classification and remote sensing image scene classification.This article conducts classification experiments for these two main directions.First,the traditional neural networks Le Net-5,Alex Net-4 and VGGNet-16 are used to conduct experiments on the natural scene classification data set,and the experimental results are analyzed to find that the experimental results of VGGNet-16 are better.Then select the integrated learning method and VGGNet-16 to construct Ada-VGG16 and RFVGG16,compare the experimental results and select RF-VGG16 to transform it into a shallow convolutional neural network RF-VGG6,and compare several convolution The experimental results of the neural network obtained the new shallow convolutional neural network RF-VGG6 based on integrated learning proposed in this paper.The classification accuracy of the experiment is good and the experiment speed is fast.It is a model suitable for natural image scene classification.Then,another important direction of scene classification is used to verify the proposed RF-VGG6 model using remote sensing scene classification data sets,and the experimental results obtained are good.It is concluded that the model may be transferable to the scene classification data set.In addition,this paper proposes a new data processing method for the characteristics of small data sets.By changing the color picture into a gray image and recoloring the method,the experimental results are obtained.The RF-VGG6 model proposed in this paper may be feasible and generalizable for future image scene classification problems.