To analyze the characteristics of the aeromagnetic scalar gradient detection method, a uniformly magnetized ellipsoid is used to simulate an unexploded ordnance, and a magnetic field detection model is established in the International Geomagnetic Reference Field based on rotation matrices. Furthermore, the spatial distribution of the target’s magnetic field is simulated. The results indicate that the scalar gradient detection curve is closely related to the unmanned aerial vehicle (UAV) heading, geomagnetic direction, and target attitude. According to the measured data, the aeromagnetic detection system exhibits differences in the detection of different headings, indicating that some “blind areas” exist in the scalar gradient magnetic detection method. The experimental measurement by a quadrotor UAV equipped with two optical pump magnetometers verifies that the scalar gradient detection method can effectively eliminate the geomagnetic field as well as the interferences of the UAV itself. Furthermore, the angular relationship between the target magnetic field contour distribution and the heading is found to be the main reason that the scalar gradient detection system enters the “blind detection area.” Therefore, a flight strategy of “positive direction + orthogonal grid” is proposed. This method effectively reduces the missed detection rate of scalar gradient detection and provides strategic guidance for the detection path of aeromagnetic scalar gradient system.

The aeromagnetic gradient detection system composed of two vector magnetic detectors carried by the Unmanned Aerial Vehicle (UAV) will be affected by error parameters such as non-orthogonality, non-alignment, sensitivity gain, and zero drift. The coupling relationship between these parameters complicates the process of solving the error parameters. Aiming at this problem, a linear calibration model of error parameters is constructed. The model establishes a system of linear equations using the projection relationship between the two vector magnetic sensors. This system of equations simplifies the process of solving the error parameters. To verify the correction effect of the model, a training process of arbitrary shaking of the global surface attitude is proposed to simulate the flight attitude of the UAV. The method has little dependence on the test equipment and has strong environmental adaptability. It can quickly solve the non-alignment error and relative zero drift error parameters on the ground. The actual measurement results show that the calibration effect between the corresponding axes can reach more than 89% under random shaking of an angle of ±20°.