An adaptive exposure algorithm for stereo imaging and its performance in an orchard

Stereo vision is being introduced in perception systems for autonomous agricultural vehicles. When working outdoors, light conditions change continuously. The perception system should be able to continuously adapt and correct camera exposure parameters to obtain the best interpretation of the scene practically possible.

We describe the development and testing of an algorithm to update exposure parameter camera setting of a stereoscopic camera under dynamic light conditions. Static tests using a stereo camera were carried out in an orchard to determine how 2D image histograms and the 3D reconstruction change with exposure.

An algorithm based on an "ideal mean pixel value" in the image was developed and implemented on the perception system of an automatic tractor. The system was tested in an orchard and found to perform satisfactorily under different orchard and light conditions.