Method for non-contact ground moisture level estimation at various depths using reflected energy at a work machine

Abstract

According to a method, model generation includes receiving inputs comprising actual soil properties comprising moisture values at a training ground surface and corresponding ground depths, temperature, surface-RF antenna distance, and antenna vibration. RF signals across a first frequency range are emitted toward the ground surface, and characteristic values determined based on reflected energy. The model is iteratively trained to observe correlations between various inputs and characteristic values at corresponding frequencies. Real-time and non-contact moisture estimation includes corresponding inputs, and a vector network analyzer identifies target frequency ranges as subsets of the first range, and directs RF emittion across the target range toward the ground. Using characteristic values determined from reflected RF signals and further reference to the trained model, moisture values for the current ground surface and corresponding depths are estimated, and an output signal generated for display, control of irrigation, planting depth, etc.