Early and accurate prediction of grain yield is of great significance for ensuring food security and formulating food policy. The exploration of key growth phases and features is beneficial to improving the efficiency and accuracy of yield prediction. In this study, a hybrid approach using the WOFOST model and deep learning was developed to forecast corn yield, which analysed yield prediction potential at different growth phases and features. The World Food Studies (WOFOST) model was used to build a comprehensive simulated dataset by inputting meteorological, soil, crop and management data. Different feature combinations at various growth phases were designed to forecast yield using machine learning and deep learning methods. The results show that the key features of corn’s vegetative growth stage and reproductive growth stage were growth state features and water-related features, respectively. With the continuous advancement of the crop growth stage, the ability to predict yield continued to improve. Especially after entering the reproductive growth stage, corn kernels begin to form, and the yield prediction performance is significantly improved. The performance of the optimal yield prediction model in flowering (R² = 0.53, RMSE = 554.84 kg/ha, MRE = 8.27%), in milk maturity (R² = 0.89, RMSE = 268.76 kg/ha, MRE = 4.01%), and in maturity (R² = 0.98, RMSE = 102.65 kg/ha, MRE = 1.53%) were given. Thus, our method improves the accuracy of yield prediction, and provides reliable analysis results for predicting yield at various growth phases, which is helpful for farmers and governments in agricultural decision making. This can also be applied to yield prediction for other crops, which is of great value to guide agricultural production.

Cotton constitutes 81% of the world’s natural fibers. Accurate and rapid cotton yield estimation is important for cotton trade and agricultural policy development. Therefore, we developed a remote sensing index that can intuitively represent cotton boll characteristics and support cotton yield estimation by extracting cotton boll pixels. In our study, the Density of open Cotton boll Pixels (DCPs) was extracted by designing different cotton boll indices combined with the threshold segmentation method. The relationship between DCP and field survey datasets, the Density of Total Cotton bolls (DTC), and yield were compared and analyzed. Five common yield estimation models, Linear Regression (LR), Support Vector Regression (SVR), Classification and Regression Trees (CART), Random Forest (RF), and K-Nearest Neighbors (KNN), were implemented and evaluated. The results showed that DCP had a strong correlation with yield, with a Pearson correlation coefficient of 0.84. The RF method exhibited the best yield estimation performance, with average R² and rRMSE values of 0.77 and 7.5%, respectively (five-fold cross-validation). This study showed that RedGreenBlue (RGB) and Near Infrared Red (NIR) normalized, a normalized form index consisting of the RGB and NIR bands, performed best.

Currently, remote sensing crop identification is mostly based on all available images acquired throughout crop growth. However, the available image and data resources in the early growth stage are limited, which makes early crop identification challenging. Different crop types have different phenological characteristics and seasonal rhythm characteristics, and their growth rates are different at different times. Therefore, making full use of crop growth characteristics to augment crop growth difference information at different times is key to early crop identification. In this study, we first calculated the differential features between different periods as new features based on images acquired during the early growth stage. Secondly, multi-temporal difference features of each period were constructed by combination, then a feature optimization method was used to obtain the optimal feature set of all possible combinations in different periods and the early key identification characteristics of different crops, as well as their stage change characteristics, were explored. Finally, the performance of classification and regression tree (Cart), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), and Support Vector Machine (SVM) classifiers in recognizing crops in different periods were analyzed. The results show that: (1) There were key differences between different crops, with rice changing significantly in period F, corn changing significantly in periods E, M, L, and H, and soybean changing significantly in periods E, M, N, and H. (2) For the early identification of rice, the land surface water index (LSWI), simple ratio index (SR), B11, and normalized difference tillage index (NDTI) contributed most, while B11, normalized difference red-edge3 (NDRE3), LSWI, the green vegetation index (VIgreen), red-edge spectral index (RESI), and normalized difference red-edge2 (NDRE2) contributed greatly to corn and soybean identification. (3) Rice could be identified as early as 13 May, with PA and UA as high as 95%. Corn and soybeans were identified as early as 7 July, with PA and UA as high as 97% and 94%, respectively. (4) With the addition of more temporal features, recognition accuracy increased. The GBDT and RF performed best in identifying the three crops in the early stage. This study demonstrates the feasibility of using crop growth difference information for early crop recognition, which can provide a new idea for early crop recognition.

Early crop identification can provide timely and valuable information for agricultural planting management departments to make reasonable and correct decisions. At present, there is still a lack of systematic summary and analysis on how to obtain real-time samples in the early stage, what the optimal feature sets are, and what level of crop identification accuracy can be achieved at different stages. First, this study generated training samples with the help of historical crop maps in 2019 and remote sensing images in 2020. Then, a feature optimization method was used to obtain the optimal features in different stages. Finally, the differences of the four classifiers in identifying crops and the variation characteristics of crop identification accuracy at different stages were analyzed. These experiments were conducted at three sites in Heilongjiang Province to evaluate the reliability of the results. The results showed that the earliest identification time of corn can be obtained in early July (the seven leaves period) with an identification accuracy up to 86%. In the early stages, its accuracy was 40~79%, which was low, and could not reach the satisfied accuracy requirements. In the middle stages, a satisfactory recognition accuracy could be achieved, and its recognition accuracy was 79~100%. The late stage had a higher recognition accuracy, which was 90~100%. The accuracy of soybeans at each stage was similar to that of corn, and the earliest identification time of soybeans could also be obtained in early July (the blooming period) with an identification accuracy up to 87%. Its accuracy in the early growth stage was 35~71%; in the middle stage, it was 69~100%; and in the late stage, it was 92~100%. Unlike corn and soybeans, the earliest identification time of rice could be obtained at the end of April (the flooding period) with an identification accuracy up to 86%. In the early stage, its accuracy was 58~100%; in the middle stage, its accuracy was 93~100%; and in the late stage, its accuracy was 96~100%. In terms of crop identification accuracy in the whole growth stage, GBDT and RF performed better than other classifiers in our three study areas. This study systematically investigated the potential of early crop recognition in Northeast China, and the results are helpful for relevant applications and decision making of crop recognition in different crop growth stages.