Scene recognition method based on convolution multi-features and deep random forest

Abstract

The invention discloses a scene recognition method based on convolution multi-features and deep random forest. A sparsely coded spatial pyramid matching method and a Fisher vector are applied to a convolutional neural network (CNN) for feature extraction and the extracted features are applied to the scene recognition method for the deep random forest to improve scene recognition accuracy. The method includes training a training image by using the CNN, performing Fisher vector encoding on the output of the last convolutional layer in the CNN, deconvolving the output of the CNN, calculating theimage feature point distribution of the output at different resolutions by using the spatial pyramid matching method to form a multi-scale spatial local feature, and then using the deep random forestfor classification, thereby improving scene recognition accuracy.

Description

technical field [0001] The invention relates to computer pattern recognition technology, in particular to a scene recognition method based on convolutional multi-features and deep random forest. Background technique [0002] In digital image and digital video data, there is a large amount of visual information, and computer vision technology is a technology that uses computers to intelligently extract and analyze the useful information contained in these visual information. With the rapid development of computer theory, applications and applications, the ability of computers to process images and videos has been greatly improved, making computational vision technology a key research direction in the fields of computer and artificial intelligence. [0003] The recognition and analysis of scene images is an important topic in the field of computer pattern recognition and an important branch of the field of image recognition. Scene recognition is also involved in the field of ...

AI Technical Summary

Problems solved by technology

However, this method also has problems: 1) The recognition of scene images is a top-down process, which needs to consider both global features and local features, and the effect of only using convolutional neural network for scene image recognition is not ideal ; 2) A large number of samples are required for training, which cannot be used for small-scale data tasks, and the training time is very long; 3) The structure of the deep neural network is very complex, heavily dependent on parameter tuning, and contains a large number of hyperparameters, which is not conducive to system stability, and due to Convolutional neural networks are not conducive to analysis due to many different choices such as convolutional layer structure

Images