Soybean in the field has a wide range of intermixed weed species and a complex distribution status, and the weed identification rate of traditional methods is low. Therefore, a weed identification method is proposed based on the optimized Faster R-CNN algorithm for the soybean seedling. Three types of weed datasets, including soybean, with a total of 9816 photos were constructed, and cell phone photo data were used for training and recognition. Firstly, by comparing the classification effects of ResNet50, VGG16, and VGG19, VGG19 was identified as the best backbone feature extraction network for model training. Secondly, an attention mechanism was embedded after the pooling layer in the second half of VGG19 to form the VGG19-CBAM structure, which solved the problem of low attention to the attention target during model training using the trained Faster R-CNN algorithm to identify soybean and weeds in the field under the natural environment and compared with two classical target detection algorithms, SSD and Yolov4. The experimental results show that the Faster R-CNN algorithm using VGG19-CBAM as the backbone feature extraction network can effectively identify soybeans and weeds in complex backgrounds. The average recognition speed for a single image is 336 ms, and the average recognition accuracy is 99.16%, which is 5.61% higher than before optimization, 2.24% higher than the SSD algorithm, and 1.24% higher than the Yolov4 algorithm. Therefore, this paper’s optimized target detection model is advantageous and can provide a scientific method for accurate identification and monitoring of grass damage.

Timely and reliable maize yield prediction is essential for the agricultural supply chain and food security. Previous studies using either climate or satellite data or both to build empirical or statistical models have prevailed for decades. However, to what extent climate and satellite data can improve yield prediction is still unknown. In addition, fertilizer information may also improve crop yield prediction, especially in regions with different fertilizer systems, such as cover crop, mineral fertilizer, or compost. Machine learning (ML) has been widely and successfully applied in crop yield prediction. Here, we attempted to predict maize yield from 1994 to 2007 at the plot scale by integrating multi-source data, including monthly climate data, satellite data (i.e., vegetation indices (VIs)), fertilizer data, and soil data to explore the accuracy of different inputs to yield prediction. The results show that incorporating all of the datasets using random forests (RF) and AB (adaptive boosting) can achieve better performances in yield prediction (R²: 0.85~0.98). In addition, the combination of VIs, climate data, and soil data (VCS) can predict maize yield more effectively than other combinations (e.g., combinations of all data and combinations of VIs and soil data). Furthermore, we also found that including different fertilizer systems had different prediction accuracies. This paper aggregates data from multiple sources and distinguishes the effects of different fertilization scenarios on crop yield predictions. In addition, the effects of different data on crop yield were analyzed in this study. Our study provides a paradigm that can be used to improve yield predictions for other crops and is an important effort that combines multi-source remotely sensed and environmental data for maize yield prediction at the plot scale and develops timely and robust methods for maize yield prediction grown under different fertilizing systems.

Aiming at the problem that it is difficult to identify two types of weeds, grass weeds and broadleaf weeds, in complex field environments, this paper proposes a semantic segmentation method with an improved UNet structure and an embedded channel attention mechanism SE module. First, to eliminate the semantic gap between low-dimensional semantic features and high-dimensional semantic features, the UNet model structure is modified according to the characteristics of different types of weeds, and the feature maps after the first five down sampling tasks are restored to the same original image through the deconvolution layer. Hence, the final feature map used for prediction is obtained by the fusion of the upsampling feature map and the feature maps containing more low-dimensional semantic information in the first five layers. In addition, ResNet34 is used as the backbone network, and the channel attention mechanism SE module is embedded to improve useful features. The channel weight is determined, noise is suppressed, soybean and grass weeds are identified, and broadleaf weeds are extracted through digital image morphological processing, and segmented images of soybean plants, grass weeds and broadleaf weeds are generated. Moreover, compared with the standard semantic segmentation models, FCN, UNet, and SegNet, the experimental results show that the overall performance of the model in this paper is the best. The average intersection ratio and average pixel recognition rate in a complex field environment are 0.9282 and 96.11%, respectively. On the basis of weed classification, the identified weeds are further refined into two types of weeds to provide technical support for intelligent precision variable weed spraying.

Research on fusion modeling of high spatial and temporal resolution images typically uses MODIS products at 500 m and 250 m resolution with Landsat images at 30 m, but the effect on results of the date of reference images and the ‘mixed pixels’ nature of moderate-resolution imaging spectroradiometer (MODIS) images are not often considered. In this study, we evaluated those effects using the flexible spatiotemporal data fusion model (FSDAF) to generate fusion images with both high spatial resolution and frequent coverage over three cotton field plots in the San Joaquin Valley of California, USA. Landsat images of different dates (day-of-year (DOY) 174, 206, and 254, representing early, middle, and end stages of the growing season, respectively) were used as reference images in fusion with two MODIS products (MOD09GA and MOD13Q1) to produce new time-series fusion images with improved temporal sampling over that provided by Landsat alone. The impact on the accuracy of yield estimation of the different Landsat reference dates, as well as the degree of mixing of the two MODIS products, were evaluated. A mixed degree index (MDI) was constructed to evaluate the accuracy and time-series fusion results of the different cotton plots, after which the different yield estimation models were compared. The results show the following: (1) there is a strong correlation (above 0.6) between cotton yield and both the Normalized Difference Vegetation Index (NDVI) from Landsat (NDVI_L30) and NDVI from the fusion of Landsat with MOD13Q1 (NDVI_F250). (2) Use of a mid-season Landsat image as reference for the fusion of MODIS imagery provides a better yield estimation, 14.73% and 17.26% higher than reference images from early or late in the season, respectively. (3) The accuracy of the yield estimation model of the three plots is different and relates to the MDI of the plots and the types of surrounding crops. These results can be used as a reference for data fusion for vegetation monitoring using remote sensing at the field scale.

Soil organic matter (SOM) is very important to the quality evaluation of cultivated land, especially in fertile black soil areas. Many studies use remote sensing images combined with different machine learning algorithms to predict the regional SOM content. However, the information provided by remote sensing images in different time windows is very different. Taking Youyi Farm, a typical black soil area in Northeast China, as the research area, this study obtains all available Sentinel-2 images covering the research area from 2019 to 2021, calculates the spectral index of single-phase and multi-temporal synthesis images, takes the spectral index and band of each image as the input, and employs the random forest regression algorithm to evaluate the performance of SOM prediction using remote sensing images with different time windows. The results show that: (1) the accuracy of SOM prediction using image band and spectral index is generally improved compared to using only the band; (2) when using single-phase images, the R² range of SOM prediction using image band and spectral index is from 0.16 to 0.59 and the RMSE ranges from 0.82% to 1.23%; When using multi-temporal synthesis images, the R² range of SOM prediction using image band and spectral index is from 0.18 to 0.56 and the RMSE ranges from 0.85% to 1.19%; (3) the highest accuracy of SOM prediction using synthetic images is lower than that of single-phase images; (4) the best time window of the bare soil period in the study area is May. This study emphasizes the importance of the time window to SOM prediction. In subsequent SOM prediction research, remote sensing images with appropriate time windows should be selected first, and then the model should be optimized.

Farmland shelterbelts are aimed at farmland protection and productivity improvement, environmental protection and ecological balance, as well as land use planning and management. Farmland shelterbelts play a vital role in determining the structural integrity and overall effectiveness of farmland, and assessing the dynamic changes within these protective forests accurately and swiftly is essential to maintaining their protective functions as well as for policy formulation and effectiveness evaluation in relevant departments. Traditional methods for extracting farmland shelterbelt information have faced significant challenges due to the large workload required and the inconsistencies in the accuracy of existing methods. For example, the existing vegetation index extraction methods often have significant errors, which remain unresolved. Therefore, developing a more efficient extraction method with greater accuracy is imperative. This study focused on Youyi Farm in Heilongjiang Province, China, utilizing satellite data with spatial resolutions ranging from 0.8 m (GF-7) to 30 m (Landsat). By taking into account the growth cycles of farmland shelterbelts and variations in crop types, the optimal temporal window for extraction is identified based on phenological analysis. The study introduced a new index—the Re-Modified Anthocyanin Reflectance Index (RMARI)—which is an improvement on existing vegetation indexes, such as the NDVI and the improved original ARI. Both the accuracy and extraction results showed significant improvements, and the feasibility of the RMARI was confirmed. The study proposed four extraction schemes for farmland shelterbelts: (1) spectral feature extraction, (2) extraction using vegetation indexes, (3) random forest extraction, and (4) RF combined with characteristic index bands. The extraction process was implemented on the GEE platform, and results from different spatial resolutions were compared. Results showed that (1) the bare soil period in May is the optimal time period for extracting farmland shelterbelts; (2) the RF method combined with characteristic index bands produces the best extraction results, effectively distinguishing shelterbelts from other land features; (3) the RMARI reduces background noise more effectively than the NDVI and ARI, resulting in more comprehensive extraction outcomes; and (4) among the satellite images analyzed—GF-7, Planet, Sentinel-2, and Landsat OLI 8—GF-7 achieves the highest extraction accuracy (with a Kappa coefficient of 0.95 and an OA of 0.97), providing the most detailed textural information. However, comprehensive analysis suggests that Sentinel-2 is more suitable for large-scale farmland shelterbelt information extraction. This study provides new approaches and technical support for periodic dynamic forestry surveys, providing valuable reference points for agricultural ecological research.