For the low resolution of thermal infrared image, the crop canopy area can not be accurately extracted, and accurate canopy temperature can not be obtained. Maize in the jointing stage was taken as research object, and the thermal infrared image and orthophoto were obtained by using the UAV equipped with thermal infrared imager and the Dajiang Elf Pro UAV. Based on highresolution orthophotos, the improved Canny edge detection operator, support vector machine (SVM) and wavelet transform were used to extract the maize canopy region, and the classification results were binarized in thermal infrared imaging. The maize canopy temperature was extracted by using the mask generated by the binarization result. The extracted vector surface analysis extraction effect was applied and the accuracy of the three extraction algorithms was evaluated. The experimental results showed that the effects of the three methods from strong to weak were as follows: improved Canny edge detection operator, SVM and wavelet transform;the extraction accuracy was 87.3%, 74.5% and 68.2%, respectively. At the same time, the error analysis of the maize canopy temperature measured by the handheld thermometer and the extracted canopy temperature was performed. The experimental results showed that the correlation between the canopy temperature extracted by the three algorithms and the measured temperature of the ground from strong to weak was as follows: 0.9295, 0.8957 and 0.8760. The improved Canny edge detection operator can better extract the maize canopy area and obtain more accurate maize canopy temperature, so as to more effectively monitor the physiological status of maize, predict drought, and formulate reasonable irrigation and fertilization measures to increase maize yield.