The invention discloses an intelligent weeding method and
system based on a
deep learning algorithm, and relates to the technical field of
laser weeding, and the method comprises the steps: data collection, analysis,
laser adjustment, synchronous collection of multispectral images, thermal imaging images and surface temperature data, and fusion of spectral reflection characteristics and
thermal radiation characteristics of
vegetation. Through analysis of the environment average temperature and the
vegetation spectral index, according to real-time illumination and temperature conditions, the contribution proportion of the spectral features and the thermal features in the fusion process is adjusted, the sensing blind area of a single sensor under strong light, shadow or dust
fog is overcome, and meanwhile in combination with
laser irradiation parameters generated based on
weed information and
land surface temperature data, the detection accuracy is improved. And the minimum and sufficient inactivation energy of each
weed is analyzed through an ecological
energy analysis formula, the balanced energy release is realized according to whether the
soil temperature exceeds an ecological protection threshold value or not, and the
energy consumption is minimized and the potential
thermal shock to the
soil ecology is reduced on the premise of ensuring the weeding effect.