Experiment Study of Real-time Recognition of Weeds in Cotton Field Based on DSP

摘要

Considering of large amount of data and complexity of calculating in recognition of weeds from cotton fields, the research on real-time recognition of weeds from cotton fields was developed based on DSP chip TMS320DM642. The structure and feature of TMS320DM642 had been analyzed. Also several image processing procedures, including image segmentation, image filter processing, division of spray area and weed recognition whose rate is 86.7% had been studied. Because of quick image processing of DSP, real-time recognition of weed in cotton fields was increased. The average time of processing an image is 976.3ms, which can make the movement speed of weeds on the platform faster than 0.2m/s. The tests showed that real-time recognition system can meet the requirements of precise spraying on target in cotton fields.