Acquiring navigation deviation in farmland by fusing satellite and visual information

Abstract. Traditional navigation method in farmland mainly adopts single navigation approach like satellite navigation or visual navigation which is susceptible to interference and inaccurate. To improve the navigation accuracy in complex farmland environment, an algorithm for acquiring navigation deviation in farmland is proposed by fusing satellite and machine vision information. The algorithm is developed based on the satellite navigation accuracy during the plough and land preparation in the early stage of agricultural production, establishes the constraint relationship between the satellite and the visual navigation reference line, and judges the validity of the satellite and visual navigation reference line in real time according to the constraint relationship. When the validity is high, the visual navigation deviation determines the final navigation deviation. When the validity is low, the confidence degree of satellite and visual navigation deviation are calculated respectively, and then a weighted computational filtering method is applied. The filtered navigation deviation is the final navigation deviation. MATLAB simulation results show that compared with the pure visual deviation information, the average deviation of declination is reduced by 7.29%, and the average deviation of offset is reduced by 4.80%. The fusion algorithm significantly improves the stability and accuracy of single navigation deviation information acquisition, and has a broad application prospect in the unstructured environment of farmland.