Indoor scene recognition method combining deep learning and sparse representation

Abstract

The invention discloses an indoor scene recognition method combining deep learning and sparse representation, which comprises the steps of randomly selecting a plurality of indoor scene images from an indoor scene library to act as a training sample, and enabling the remaining indoor scene images to act as a test sample; performing object category discrimination and detection on the training sample and the test sample by using a Fast-RCNN algorithm so as to build low-level features of each indoor scene image; combining the low-level features and spatial features of each indoor scene image by using a bag-of-words model so as to build middle-level features; mixing the middle-level features of the training sample so as to build a sparse dictionary; performing sparse representation on the test sample by using the sparse dictionary, calculating a residual error according to a solved sparse solution and the inputted test sample, and judging an object category to which the test sample belongs according to the residual error; and outputting the judged object category to which the test sample belongs. The indoor scene recognition method can accurately recognize an indoor scene, can effectively improve the accuracy and the robustness of indoor scene recognition and has very high practical performance.

Description

technical field [0001] The invention relates to an indoor scene recognition method combined with deep learning and sparse representation, and belongs to the technical field of image processing technology. Background technique [0002] With the development and popularization of information technology and intelligent robots, scene recognition, as an important research content, has become an important research issue in the field of computer vision and pattern recognition. Scene image classification is the automatic classification of image datasets according to a given set of semantic labels. The scene recognition model is mainly divided into three major blocks: based on low-level features, based on mid-level features, and based on visual vocabulary. The so-called low-level features are to extract the global or block texture, color and other features of the scene image to classify the scene image, such as the research of Valiaya and Szumme et al., but this method of extracting ...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art, provide an indoor scene recognition method that combines deep learning and sparse representation, and solve the problems of small differences, occlusion, scale, and angle changes due to the current indoor scene intra-category differences. The current indoor scene recognition is more complex and difficult than the outdoor scene recognition, so the recognition effect is poor, so as to improve the recognition rate and robustness of the indoor scene recognition algorithm.

Images