An integrated Micro Attitude Heading Reference System(MAHRS) based on Micro-electro-mechanical System(MEMS) gyroscopes, accelerometers, and magnetic-sensors was researched for Micro Aerial Vehicles (MAV). As traditional algorithms could not keep the attitude accuracy of the MAV during a long-time maneuvering, an adaptive attitude estimation filtering algorithm with low-cost multi-sensors was presented for the MAHRS. Firstly, this algorithm was used to establish the models of acceleration and magnetic-disturbance and took them into a state equation. In this way, it could maintain the attitude accuracy of loads during the long-time maneuvering. Then, this algorithm adopted the federated filter mode to reduce the interaction between acceleration and magnetic disturbance to improve its precision and reliability. Moreover, the algorithm made the estimation variance (P) and measurement variance (R) remain steady adaptively during different maneuvering states. By comparing with other different algorithms for the MAHRS, It shows that the presented algorithm is more accurate and reliable than other algorithms during various maneuvering states.