Vision-based adaptive variable rate spraying approach for unmanned aerial vehicles

The rapid developments of unmanned aerial vehicles (UAV) and vision sensor are contributing a great reformation in precision agriculture. Farmers can fly their UAV spraying pesticides around their crop fields while staying at their remote control room or any place that is separated from their farm land. However, there is a common phenomenon in rice planting management stage that some empty areas are randomly located in farmland. Therefore, a critical problem is that the waste of pesticides that occurs when spraying pesticides over rice fields with empty areas by using the common UAV, because it is difficult to control the flow accuracy based on the empty areas changing. To tackle this problem, a novel vision-based spraying system was proposed that can identify empty areas automatically while spraying a precise amount of pesticides on the target regions. By this approach, the image was preprocessed with the Lucy-Richardson algorithm, then the target area was split from the background with k-means and the feature parameters were extracted, finally the feature parameters were filtered out with a positive contribution which would serve as the input parameters of the support vector machine (SVM) to identify the target area. Also a fuzzy control model was analyzed and exerted to compensate the nonlinearity and hysteresis of the variable rate spraying system. Experimental results proved that the approach was applicable to reducing the amount of pesticides during UAV spraying, which can provide a reference for precision agriculture aviation in the future.