The invention relates to a visible light remote sensing image candidate target extraction and classification method, belongs to the technical field of remote sensing image processing and analysis, andsolves the problems of too many candidate targets extracted by the existing method and low classification precision. The method comprises the following steps: extracting a large-scale region in a remote sensing image through a first sliding window, and inputting the large-scale region into a trained candidate region recognition model to obtain a candidate region containing an interested target; extracting a candidate target in the candidate region through a second sliding window; and performing two-step classification on the extracted candidate targets by using the trained candidate target classification model, and determining the categories of the candidate targets. According to the method, under the same recall rate, the number of the extracted candidate targets is far smaller than thatof the candidate targets obtained through an existing method, and meanwhile under the condition that the same feature and the same classifier are used, through two-step classification, the candidatetarget classification precision is greatly improved.