Temperature Drift and Compensation in Stereo Visual Odometry

In visual navigation, image sensors need to operate continuously, and the heat generated by the electronic components will cause the camera’s temperature to rise. This phenomenon in temperature affects the refractive index of the camera lens and the mechanical structure of the cameras, subsequently altering the projected coordinates of spatial points on the imaging plane. Consequently, this temperature-induced distortion impacts the accuracy of vision navigation calculations. This study addresses the challenge of camera self-heating in stereo vision navigation. It involves an analysis of the temperature error model, introduces the concept of the temperature distortion coefficient, and proposes a method for measuring and compensating for temperature distortion based on the model. Simulations and experiments demonstrate that the self-heating of cameras results in an approximate 2‰ increase in internal parameters, causing a corresponding 2‰ error in the final results. Then, by implementing temperature distortion compensation, the calculation error in stereo visual odometry caused by temperature dropped by about 80% in our experiments. Importantly, this compensation method does not impose an additional burden on computing resources during visual navigation calculations. Therefore, considering the interference caused by camera self-heating in the vision navigation process can substantially enhance navigation accuracy.