To address the cost issue associated with pixel-level image annotation in fully supervised semantic segmentation, a method based on semi-supervised semantic segmentation is proposed for extracting winter wheat planting areas. This approach utilizes self-training with pseudo-labels to learn from a small set of images with pixel-level annotations and a large set of unlabeled images, thereby achieving the extraction. In the constructed initial dataset, a random sampling strategy is employed to select 1/16, 1/8, 1/4, and 1/2 proportions of labeled data. Furthermore, in conjunction with the concept of consistency regularization, strong data augmentation techniques are applied to the unlabeled images, surpassing classical methods such as cropping and rotation to construct a semi-supervised model. This effectively alleviates overfitting caused by noisy labels. By comparing the prediction results of different proportions of labeled data using SegNet, DeepLabv3+, and U-Net, it is determined that the U-Net network model yields the best extraction performance. Moreover, the evaluation metrics MPA and MIoU demonstrate varying degrees of improvement for semi-supervised semantic segmentation compared to fully supervised semantic segmentation. Notably, the U-Net model trained with 1/16 labeled data outperforms the models trained with 1/8, 1/4, and 1/2 labeled data, achieving MPA and MIoU scores of 81.63%, 73.31%, 82.50%, and 76.01%, respectively. This method provides valuable insights for extracting winter wheat planting areas in scenarios with limited labeled data.

Effective monitoring of surface damage in mining areas is vital for ecological restoration. Unmanned aerial vehicles (UAVs) have been widely used to obtain ground subsidence data owing to their low cost and ease of operation. The images captured by UAVs can generate dense image matching (DIM) point clouds, which, after screening, can be used to create a digital elevation model (DEM) required for deformation analysis. Existing filtering algorithms mainly rely on the spatial geometric features of point clouds and rarely utilize color information, which limits their accuracy in areas with vegetation coverage. To address this issue, this study proposes a H-PTD method that combines visible light with progressive triangulated irregular network densification (PTD). First, initial ground seeds are selected based on the H value in the HSV space. Subsequently, a triangulated irregular network (TIN) is constructed, and iterative densification is performed by evaluating the relationship between the target point and adjacent triangular faces, thereby achieving an accurate distinction between ground and non-ground. Evaluated on three terrain datasets and against five classical methods, the results indicate that the Total error in the H-PTD cross-matrix is controlled between 2.9% and 7.8%, and remains below 8% overall. The standard deviation of the DEM difference is around 0.02 m. Compared to other methods, H-PTD shows higher filtering accuracy and better terrain adaptability, making it more promising for monitoring mining areas and providing a more reliable tool for subsidence detection based on UAVs.