The invention relates to an ultra-wideband non-line-of-sight identification method based on random forest. The method comprises the steps that (1) ultra-wideband data in different environments are collected to acquire original data; category labels are labelled according to the real distance information; a number of data are randomly selected as a training set and a test set; (2) feature reconstruction is carried out based on the measured ultra-wideband signal features; (3) the reconstructed features are randomly selected to build a bisected recursive decision tree; (4) samples in the training set in the step (1) are selected and put back to build a CART decision tree model; and (5) M groups of data are extracted from the training set as training samples and are put back; the CART decision tree model built in the step (4) is used for classification determining; the process is repeated for N times to form a random forest model, wherein N is greater than or equal to 30. Whether the ultra-wideband data are non-line-of-sight can be effectively identified online in real time. The practical effect is great.