When the high altitude unmanned aerial vehicle is extinguishing, it is necessary to detect the fire point. The traditional detection method is easily disturbed by the environment, its reaction is insensitive and its real-time feature is poor. Therefore, a new method based on binocular stereo vision is proposed in this study. First, we set up the experiment platform and design the hardware circuit. Then the classical binocular stereo vision ranging algorithm is studied deeply, and a new method of distance measurement is proposed. The method mainly includes:image acquisition, region of interest extraction, image preprocessing, and dynamic template matching based on the weight fusion. Finally, the multi segment nonlinear compensation method is used to construct the distance measurement model, and a large number of experiments are carried out on the experimental platform. The experimental result show that within 50 meters, the error of the ranging model is better than 1 meter, and within 100 meters, the ranging error is better than 2 meters, the matching precision is 6.947 pixels/m. The result basically meets the requirement of precise ranging for high altitude UAV fire extinguishing, so the distance measurement accuracy of the ranging model and the engineering application are both good.