Influence of images recording height and crop growth stage on leaf cover estimates and their performance in yield prediction models

Abstract

Field experiments were conducted in 1996 and 1997 to study the effects of crop growth stage and images recording height (size of the area included in the image) on the estimates of leaf cover obtained through digital image analysis techniques, and to test the effectiveness of these data in maize yield prediction. Average maize leaf cover estimates were highly variable and the precision was influenced by both the crop growth stage and the camera recording height. Sampling images from 3.3 m above the ground (1.63 m 2/image) produced less variable estimates than when sampling was done at lower heights (smaller area per image). Early timing of leaf cover assessment produced estimates with smaller variability than later timings. Maize yield prediction was slightly affected by the timing of leaf cover sampling. However, sampling at the six-leaf stage of maize may be more appropriate from a practical stand point than earlier or later samplings. While being compatible with many control options, this would allow enough weed seeds to germinate. Increasing the images recording height and consequently the area included in the images improved the accuracy of yield predictions. Data from images taken at 3.3 m fitted the model better than those from images recorded at 1.5 and 1.9 m. These results indicate that appropriate timing of leaf cover assessment and appropriate selection of image shooting height may help improve the accuracy of crop yield prediction, and thereby, reduce the risk of making wrong weed control decisions.