The invention discloses a weakly supervised fine-grained object classification method, which comprises the following steps: for an image to be classified, generating a series of rectangular local areas through an area suggestion network, sorting the local areas according to a descending order of confidence coefficients, and selecting M local areas; extracting feature vectors of the M local areas through a feature extractor so as to predict probability values of the M local areas for each category; wherein a prediction result is used for a subsequent integrated prediction part and optimizationof a regional suggestion network and a classifier in integrated prediction; extracting a feature vector of the to-be-classified image through a feature extractor, thereby predicting a probability value of the to-be-classified image as each category; and calculating a final probability value of each category by combining the probability values of the first K local areas in the M local areas as eachcategory and the probability values of the to-be-classified images as each category, and selecting the category with the maximum probability value as a classification result. The method has the advantages of being high in classification accuracy and low in cost.