Aiming at the practical problems such as insufficient accuracy and poor robustness when a single sensor is used for UAV attitude estimation, this paper proposes a multi-sensor fusion algorithm based on the Extended Kalman Filter (EKF). With the Inertial Measurement Unit (IMU) as the core sensing component, the scheme fuses Global Positioning System (GPS) and magnetometer data to construct a 16-dimensional state space model, realizing the joint solution of UAV attitude, velocity, and position information. Experimental verification shows that in static tests, the Root Mean Square Error (RMSE) of pitch and roll angle estimation errors is significantly reduced; in dynamic test scenarios, the attitude tracking accuracy is significantly improved compared with the traditional complementary filter algorithm. On this basis, this paper further builds a modular experimental platform for vocational education practice, decomposing the algorithm implementation process into four core teaching modules: sensor data collection, error calibration, state prediction, and measurement update. It provides reusable teaching cases and a complete engineering implementation idea for the curriculum reform of the UAV application technology major in higher vocational colleges.