Maize Plant Recognition Technology Based On UAV Images

With the increasing global population base and the increasing demand for food,the problem of food supplying is becoming more and more serious.Maize is one of the main crops in the world that we need to pay close attention to in the future.The application of advanced information technology in maize production will become one of the important means to improve food production efficiency.In the past,maize plant identification technology was studied with high-definition cameras,pictures were taken manually.Sample experiments were taken to evaluate the inefficiency of large farmland.There would be some limitations and inaccuracies.Now with the use of UAV,we can take large-scale photographs and samples,and view the growth status of maize plants in the whole field through aerial photographs.Through image processing technology,we can estimate the growth trend of maize and the yield of Maize in advance.In this study,the DJI-Wu 2 drone was used with the RGB visible light camera to collect images of various growth stages of corn in the Wenxian experimental field in Henan Province.Firstly,maize plant and soil were segmented using multi-threshold segmentation through color space LAB:extracting the characteristics of maize vegetation,denoising through median filter,and calculating the accuracy of maize vegetation segmentation by comparing with ground truth data,the final segmentation accuracy reaches 99%.Futher more,in this study,corn stamen contour was segmented from August maize image through H channel in HSV color space,and then corn stamen feature was segmented using super-pixel segmentation.Corn stamen contour feature was extracted by connected region,and then counted automatically by connected region.Compared with manual counting,the accuracy is as high as 96%.This study focuses on providing technical support on monitoring and analyzing the growth status of maize and can also for estimating the yield of maize.