Training sample distribution method and system for multi-scale target detection

The invention relates to a training sample distribution method and system for multi-scale target detection, and the method comprises the steps: obtaining a target data set, extracting a plurality of feature maps from the target data set through a convolutional neural network, and determining a prior frame and a prediction frame; calculating a first overlapping degree value of the prior frame and the real labeling frame and a second overlapping degree value of the prediction frame and the real labeling frame, taking the prior frame of which the first overlapping degree value is greater than a first set threshold value as a first candidate positive sample, and taking the prediction frame of which the second overlapping degree value is greater than the first set threshold value as a supplementary first candidate positive sample; constructing a first positive sample set, and calculating a mean value and a variance to obtain a second set threshold value; dividing the prior frame according to the relation between t