The invention provides a field agricultural production scene perceptual computing method and device. The method comprises the steps that binocular image data and growth environment data of crops are collected; point cloud data processing is carried out according to the collected binocular image data, and the plant height, the plant spacing and the row spacing of the crops are computed; digital image processing is carried out according to the collected binocular image data, and the canopy density and growth vigor information of the crops are computed; the plant height, the plant spacing, the row spacing, the canopy density and the growth vigor obtained by computation serve as the form structure indexes of the crops, and a crop growth model is optimized in real time; according to the growth environment data, the optimized crop growth model is combined, and the growth state of the organs of the crops is computed. The method reasonably integrates the computer graphic image, vision and automation technology, the form structure indexes of the crops are detected and serve as auxiliary analysis parameters, the yield of field agricultural production is forecasted, the completeness degree of production and management auxiliary decision is increased, and therefore the management efficiency of field agricultural production is improved.