With the aim to solve the problems of low identification accuracy of the front tractor, relative positioning difficulty, and difficulty in ensuring the safety of autonomous operation in the multimachine coordinated navigation operation, a method of tractor identification and positioning based on depth image and neural network was proposed. The tractor features were recognized and extracted by establishing YOLO-ZED neural network recognition model. The ZED camera was used to collect 1100 tractor images at different angles, distances, and resolutions in cloudy and sunny days, and the LabelImg marking tool was used to manually mark the collected tractor images, marking the cab as the identification target. The tractor positioning model based on the depth image was established and the binocular positioning principle was used to calculate the spatial position coordinates of the tractor relative to the machine. A fixedpoint identification and positioning test was performed on a small power tractor, and the identification and positioning results of the tractor were measured along the longitudinal, width and Scurve directions of the tractor. The test results showed that the algorithm can quickly and accurately identify and locate the spatial position of the tractor, and the average identification and positioning speed was 19f/s. The maximum absolute error of positioning the tractor in the camera depth direction and width direction was 0.720m and 0.090m, respectively, the maximum relative error was 7.48% and 8.00%, and the standard deviation was less than 0.030m. The accuracy and speed requirements of tractor target identification for multimachine coordinated navigation can be met.