Indoor scene image classification method facing service robot

The invention discloses an indoor scene image classification method facing a service robot and belongs to the technical field of scene classification. The indoor scene image classification method disclosed by the invention comprises the following steps: step 1: selecting a training sample from a prior scene image library; step 2: pre-processing the training sample; step 3P training a sparse automatic encoding (SAE) model based on maximum divisibility (MSD); step 4: extracting characteristics of the training sample to obtain characteristic vectors; step 5: carrying out dimensionality reduction by using mean value pooling; step 6: combining a genetic algorithm-particle swarm optimization (GA-PSO) to train a support vector machine (SVM) to obtain parameters; and step 7: classifying newly-acquired scene images by using trained SAE model and SVM. With the adoption of the indoor scene image classification method, scenes in a complicated indoor environment can be classified so that the service robot can provide more and more accurate services according to the acquired scene images.