Evaluation of an online approach for determination of percentage residue cover

An online sensor for measuring residue cover with the help of fast image analysis offers new advantages. A prototype of such a sensor system was tested, though its quality of work under real world conditions can only be specified relative to other methods acting as a reference. Therefore, a combination of different tests was conducted. In field trials with defined surface residue densities (kg ha−1) the readings of the camera sensor, a visual standard method (grid) and visual estimations were correlated with each other and with the quantities of surface straw and tillage depth. Correlation coefficients between the methods of 0.7–0.9 were found. The methods were also checked against the results of manually controlled classification of images taken at 70 selected geo-referenced reading points. This yielded the best result for the camera sensor with a correlation coefficient of 0.962. Additionally, in a laboratory test with well-defined artificial cover rates, the characteristic behaviour of image analysis and the grid method was investigated. A slight overrating of the cover by the grid method and a small error caused by the camera lens aberration could be identified. The suitability in principle of online image processing as a new reference method has been demonstrated. However, in the field, the camera sensor provided the greatest reliability, compared to untrained human observers at cover rates below about 50%. At higher residue covers, human perception might have some advantage.

► Correlation of sensor with visual methods for residue cover was 0.7–0.9. ► Camera sensor more reliable for cover rates below about 50%. ► The approach has potential to be a rapid new reference method.