The invention relates to an image processing technology, in particular to a low-resolution pedestrian detection method and system and a storage medium. The method comprises a training process and a testing process, wherein the training process firstly determines parameters of a training set and the training process; Secondly, sequentially inputting the pictures according to the batch processing size, extracting multi-scale features of the training pictures, and reconstructing and enhancing shallow features to form a new multi-scale detection framework; And finally, carrying out frame classification and position regression, calculating training loss, carrying out back propagation, and updating weight parameters. The test process comprises the following steps: firstly, determining a test set, taking a model obtained in the training process as a test model of an algorithm, sequentially inputting test pictures in small batches, extracting multi-scale features, reconstructing and enhancingshallow features, and then classifying frames and carrying out position regression. According to the method, the deep learning network is adopted to reconstruct the shallow features, and meanwhile, the effectiveness of the shallow features is improved, so that the detection capability on low-resolution pedestrians is enhanced.