Abstract: Timely and accurate yield estimation is of great significance to agricultural management and macro decision-making. Understanding the growth trend and yield of cotton is helpful not only for farmers to make production plans, but also for foreign trade and import plans of cotton. UAV(Unmanned Aerial Vehicle) remote sensing has the characteristics of high spatial resolution, fast data acquisition, simple operation and low cost, and can quickly carry out image acquisition for a certain area, which can obtain more accurate crop distribution information. It has become an important supplement to aerial remote sensing and satellite remote sensing, and has been widely used in crop monitoring and yield estimation. However, the accuracy of yield estimation needs to be improved due to the influence of atmospheric effect and crop phenology. Based on this, a SEGT(Seedling Emergence and Growth Trend) cotton yield estimation model combining the emergence and growth trend of cotton based on the high-resolution image of UAV was constructed in this paper. Firstly, Dapeng CW-10 UAV platform with Canon camera was used to acquire high-resolution visible light images in the research areas on May 23, 2018, and the seedling emergence information of cotton in the study areas was obtained by combining vegetation index calculated with OTSU(Nobuyuki Otsu) and morphological filtering. Secondly, Parrot Sequoia sensor on the eBee SQ UAV platform was used to obtain the UAV multi-spectral images of 10 periods and the NDVI(Normalized Difference Vegetation Index) of each period was calculated. By calculating the correlation between NDVI of each period and the actual yield of cotton ,the CNDVI (Comprehensive Normalized Difference Vegetation Index) was constructed according to the grading evaluation of cotton growth state and the boll numbers of cotton per plant. Finally, combined with a single boll mass, a SEGT model was constructed to estimate the yield of cotton, and the accuracy was verified according to the measured yield data. The total area of the study areas was about 42.47 hm2, including 3 sample plots with the area of 17.72, 15.35 and 9.4 hm2 respectively. Within the three sample plots, 60 sample plots of 3 m × 3 m were evenly divided, including 40 experimental areas and 20 verification sample areas. The field measured data included the actual emergence information and actual yield of cotton The results showed that ExG-ExR (Excess Green-Excess Red) vegetation index had the best effect in recognition and extraction of cotton seedlings. The accuracy rate, recall rate and F1 value reached 93%, 92.33% and 92.66%, respectively. The extraction results of cotton seedling with VDVI(Visible-Band Difference Vegetation Index) vegetation index were relatively good, the accuracy rate, recall rate and F1 value reached 86.67%, 87.67% and 88.12%, respectively. The predicted total cotton yield of the whole study area was 261,200.75 kg, the estimated cotton yield of the three experimental plots was 107,704.32, 98 032.27 and 55 464.16 kg respectively, and the actual total yield was 101 542.08, 90 389.69 and 61 626.85 kg, the accuracy of yield estimation of three plots in the study area was 94.28%, 92.2% and 90%, respectively. The validation results showed that the determination coefficient of the model was 0.92, the root mean square error was 0.1, and the relative error was 3.47%. SEGT model had high accuracy and credibility. The study can provide references for the application of UAV remote sensing in crop yield estimation and provide new research ideas for cotton yield estimation.