Vision-aided inertial navigation for planetary landing without feature extraction and matching

In this paper, a vision-aided inertial navigation method for planetary landing based on the direct method is proposed. This approach requires only a monocular camera and an inertial measurement unit (IMU) to be equipped on the lander. Unlike traditional feature-based visual-aided inertial navigation algorithms, which entail extracting and matching image features and thus incurring high computational costs, the proposed method utilizes the direct method that tracks any pixel with a sufficient intensity gradient. By jointly optimizing photometric and inertial errors, the proposed algorithm estimates the camera motion and sparse map of the scene. Simulation results demonstrate that the proposed method reduces computation by nearly 65 % compared to the feature point method, with a slight decrease in navigation accuracy.