This paper introduces an outline of a workstation for the agricultural autonomous vehicles. It also describes a background estimation method based on the Kalman filter for detecting an agricultural autonomous vehicle. The detecting system has been achieved in eight steps: generation of a background image and initialization, adaptation of a background image, detection of a foreground image, computation for the connected components of a region; computation for the intersection of two regions; computation for regions with the aid of shape features; transformation of the shape of a region, update of a background image. In this research, performance of the background estimation method based on the Kalman filter has been compared with that of general background subtraction in terms of accuracy, using both images captured on a sunny and cloudy day. The results showed that increasing the accuracy requires more than 10 frames per second to detect an agricultural vehicle in the above conditions.