Abstract:With the rapid development of remote sensing platform of low altitude unmanned aerial vehicle (UAV), the dynamic, fast and inexpensive features, the UAV remote sensing platform has more research in various fields, especially in the precision agriculture irrigation technology. The unmanned aerial vehicle thermal infrared low altitude remote sensing technology can quickly monitor the canopy temperature information of the crop, which can further use the canopy temperature information to diagnose the water stress condition of the crop. However, the processing of high resolution thermal infrared image of UAV is the key to the diagnosis of crop moisture, eliminating the soil background of UAV thermal infrared image is an effective way to improve the accuracy of crop water diagnosis. However, it is also a difficult problem in thermal infrared image processing. Different water treatments were carried out, including I1 (50% of field holding water), I2 (65% of field holding water), I3 (80% of field holding water) and I4 (control group 95%~100% of field holding water), and each water treatment set three repeat tests, a total of 12 plots, each plot was 4m×5m）. Flower boll cotton was taken as study object at 09:00, 13:00 and 17:00 of day, respectively, and UAV high resolution thermal infrared images were obtained. Firstly, the two-valued Ostu algorithm and the Canny edge detection algorithm were used to deal with the thermal infrared image, and achieve the elimination of soil background, then, the two-valued Ostu algorithm and Canny edge detection algorithm contained soil background were used to calculate the crop water stress index, finally, the relationship models between three kinds of crop water stress index (CWSI) and cotton leaf stomatal conductance at different times was established. The researchresults showed that the application of Canny edge detection algorithm can effectively eliminate the soil background in thermal infrared image, and there was a single peak distribution of the temperature histogram after removing the soil background. Among the crop water stress index obtained from three kinds of treatment methods, the CWSI of Canny edge detection algorithm was the minimal, the CWSI of two-valued Ostu algorithm was higher, and the CWSI with soil background was the largest. The determination coefficient between CWSI and cotton leaf stomatal conductance by using Canny edge detection algorithm to remove the soil background was up to 0.84, and that by using two-valued Ostu algorithm resulted the second, and that got by containing soil background was the worst. The research result can provide a reference method for monitoring water condition of the plant by UAV thermal infrared technology.
