The invention provides a binocular distance measurement method based on an improved quadtree ORB algorithm. The method comprises the following steps: firstly, fitting a relationship between depth and parallax through a calibration plate; images are collected and preprocessed; thirdly, simplifying a retrieval domain through a quadtree decomposition algorithm, and removing a background region; feature point detection is carried out by adopting a FAST algorithm and a Kitchen-Rosenfeld algorithm at the same time, and feature points are subdivided by adopting a quadtree homogenization algorithm; matching feature points of the left photo and the right photo by using BRIEF feature description, and removing wrong matching by using a Ransac algorithm; and finally, recovering feature point coordinates through a triangulation principle, and calculating the size of the object according to a spatial distance formula. According to the method, the improved quadtree ORB matching algorithm is applied to binocular distance measurement, and the three-dimensional coordinates of the feature points are calculated by using a method for fitting the K value of the parallax and distance relationship, so that the detection efficiency and the detection precision are improved.