Plant height is one of the most critical agronomic traits to reflect the growth status, health condition, and overall vigor of the crops. Accurate and efficient measurement of the plant height is also essential for the crop monitoring and yield estimation in precision agriculture. Conventional manual measurements on the plant height are often time-consuming, labor-intensive, and prone to human error. Recent advances in the computer vision and deep learning can be expected for the non-destructive measurements on the plant phenotyping. Advanced feature extraction, attention mechanisms, and depth-to-height conversion can also be integrated to provide the high precise estimation of the plant height. In this study, a monocular height regression method (MHRM) was proposed to estimate the height of the winter wheat using single-camera images. The RGB images were also captured in the field conditions as the input. A crop target detection module was firstly applied to locate the relevant plant regions. The MHRM was effectively reduced the interference from the soil, background vegetation, and the non-crop objects. The local crop regions were then fed into a refined feature depth network, which consisted of a feature extraction, a feature refinement and a depth prediction module. Among them, the feature extraction module was combined the convolutional neural networks with the channel attention mechanisms, in order to enhance the representational capacity of the plant features. The feature refinement module was further improved the feature quality using multi-scale convolutions, depthwise separable convolutions, and efficient channel attention mechanisms, thereby enhancing the robustness of the depth estimation under varying illumination and background. Finally, the depth prediction module was utilized to generate the pixel-level depth maps. Subsequently, the real-world plant heights were converted after height generation. Furthermore, a joint supervision was employed to incorporate both pixel-level reconstruction and scale-consistency loss during training. The dual-loss configuration was improved the precision of the depth estimation. The extracted values of the plant height were well consistent with the real-world measurements. A field experiment was conducted to evaluate the performance at the Shandong Taian Agricultural Meteorological Experimental Station. Winter wheat images were collected under natural lighting and field conditions. Four representative models of the monocular depth estimation were selected as the baselines: Boosting Monocular Depth Estimation with Local Planar Guidance (BTS), Fully Convolutional Residual Network (FCRN), Deep Ordinal Regression Network (DORN), and Dense Prediction Transformer (DPT). Quantitative results indicated that the refined feature depth network was achieved in the superior performance with the high robustness and applicability, compared with all baseline models. Specifically, there were the lower root mean square error (2.759), logarithmic root mean square error (0.157), relative error (0.152), and squared relative error (0.907). Subsequently, the estimated depth maps were transformed into the plant height measurements. A comparison was then made with manually collected ground-truth data. The method was achieved in an extraction accuracy of 98.74%, thereby outperforming BTS (92.68%), FCRN (97.17%), DORN (97.44%), and DPT (98.40%). The results demonstrated that the MHRM was reliably captured the crop height information, even in the complex field conditions. In conclusion, the monocular depth estimation with the attention-enhanced feature extraction can provide an accurate, efficient, and non-destructive solution for the winter wheat height measurement. The promising potential can also offer for the broader applications in precision agriculture, crop phenotyping, and field monitoring. A practical tool can be expected for the decision-making on the high productivity of the crops in modern agriculture. The finding can also highlight the attention mechanisms and multi-scale feature refinement for the depth prediction in agricultural scenarios.

Farmland parcel data can be accurately and rapidly acquired using remote sensing in modern agriculture. In this study, a prior knowledge and semantic features of integration-based farmland parcel extraction (PKFFPE) was proposed to accurately extract the winter wheat data from the high-resolution remote sensing images. The high-resolution remote sensing images were taken as the data source. Three modules were divided into the PKFFPE: an initial segmentation, a post-processing and an extraction module. Among them, the initial segmentation module was used to construct and then train a convolutional neural network model for the pixel-by-pixel classification. The post-processing module was to optimize the initial segmentation. The extraction module was to accurately extract the farmland parcel data from the inputted remote sensing images. The initial segmentation module of PKFFPE consisted of three parts: input, feature extractor and classifier. The input included the remote sensing and edge image blocks. It was also necessary to add the pixel-by-pixel labels into the input in the training stage; The feature extractor was employed an encoder-decoder structure. Two units of feature extraction were utilized to extract the semantic and edge features. The semantic feature extraction employed a multi-scale attention mechanism, including an improved channel and spatial attention mechanism. A multi-scale feature extraction was used to capture the features at different scales. The decoder shared the six levels of decoding units, each of which contained a number of convolutional layers. The columns and rows of the feature map were recovered, according to the inputted remote sensing image block each time. A SoftMax model was used as the classifier to classify each pixel, according to the feature map output from the decoder. Each pixel was labeled as either a winter wheat or a non-winter wheat. According to the prior knowledge derived from planting management, the pixels within the same farmland parcel generally exhibited the high consistency in the basic features, such as color and texture. While there were typically greater differences in these features, compared with the adjacent objects, such as the roads. The prior knowledge was obtained from the distribution patterns of color and texture in the same farmland parcel. The optimization was performed on the initial segmentation after post-processing, in order to generate the accurate farmland parcel data. The applicability of the PKFFPE was verified in the plain and hilly areas. The study areas were taken from Guantao County in Handan City, Hebei Province, and Ningyang County in Tai'an City, Shandong Province, China. The UNet, ErfNet, SegNet, EIGNet, and object-oriented classification models were selected to compare the initial segmentation. While conditional random fields and mathematical morphology were selected to evaluate the post-processing. The experimental results show that the accuracy (96.1%, 92.3%), precision (90.6%, 87.6%), recall (93.2%, 90.6%), and F1 coefficient (91.9%, 89.0%) of the PKFFPE in Guantao and Ningyang counties were better than the rest. The PKFFPE performed the best to accurately extract the farmland parcel data from the high-resolution remote sensing images. The findings can also be applied to scientific research and practical production