Convolutional neural networks (CNNs) can adapt to more complex data, extract deeper characteristics from images, and achieve higher classification accuracy in remote sensing image scene classification and object detection compared to traditional shallow-model methods. However, directly applying common-structure CNNs to pixel-based remote sensing image classification will lead to boundary or outline distortions of the land cover and consumes enormous computation time in the image classification stage. To solve this problem, we propose a high-resolution remote sensing image classification method based on CNN and the restricted conditional random field algorithm (CNN-RCRF). CNN-RCRF adopts CNN superpixel classification instead of pixel-based classification and uses the restricted conditional random field algorithm (RCRF) to refine the superpixel result image into a pixel-based result. The proposed method not only takes advantage of the classification ability of CNNs but can also avoid boundary or outline distortions of the land cover and greatly reduce computation time in classifying images. The effectiveness of the proposed method is tested with two high-resolution remote sensing images, and the experimental results show that the CNN-RCRF outperforms the existing traditional methods in terms of overall accuracy, and CNN-RCRF’s computation time is much less than that of traditional pixel-based deep-model methods.

Since the result images obtained by deep semantic segmentation neural networks are usually not perfect, especially at object borders, the conditional random field (CRF) method is frequently utilized in the result post-processing stage to obtain the corrected classification result image. The CRF method has achieved many successes in the field of computer vision, but when it is applied to remote sensing images, overcorrection phenomena may occur. This paper proposes an end-to-end and localized post-processing method (ELP) to correct the result images of high-resolution remote sensing image classification methods. ELP has two advantages. (1) End-to-end evaluation: ELP can identify which locations of the result image are highly suspected of having errors without requiring samples. This characteristic allows ELP to be adapted to an end-to-end classification process. (2) Localization: Based on the suspect areas, ELP limits the CRF analysis and update area to a small range and controls the iteration termination condition. This characteristic avoids the overcorrections caused by the global processing of the CRF. In the experiments, ELP is used to correct the classification results obtained by various deep semantic segmentation neural networks. Compared with traditional methods, the proposed method more effectively corrects the classification result and improves classification accuracy.

Descriptions of the spatial locations of disappeared objects are often recorded in eyewitness records, travel notes, and historical documents. However, in geographic information system (GIS), the observer-centered and vague nature of the descriptions causes difficulties in representing the spatial characters of these objects. To address this problem, this paper proposes a Fuzzy Spatial Region Extraction Model for Object’s Vague Location Description from Observer Perspective (FSREM-OP). In this model, the spatial relationship between the observer and the object are represented in spatial knowledge. It is composed of “phrase” and “region”. Based on the spatial knowledge, three components of spatial inference are constructed: Spatial Entities (SEs), Fuzzy Spatial Regions (FSRs), and Spatial Actions (SAs). Through spatial knowledge and the components of FSREM-OP, an object’s location can be inferred from an observer’s describing text, transforming the vagueness and subjectivity of location description into fuzzy spatial regions in the GIS. The FSREM-OP was tested by constructing a group of observers, object position relationships and vague descriptions. The results show that it is capable of extracting the spatial information and presenting location descriptions in the GIS, despite the vagueness and subjective spatial relation expressions in the descriptions.

Many documents contain vague location descriptions of observed objects. To represent location information in geographic information systems (GISs), these vague location descriptions need to be transformed into representable fuzzy spatial regions, and knowledge about the location descriptions of observer-to-object spatial relations must serve as the basis for this transformation process. However, a location description from the observer perspective is not a specific fuzzy function, but comes from a subjective viewpoint, which will be different for different individuals, making the corresponding knowledge difficult to represent or obtain. To extract spatial knowledge from such subjective descriptions, this research proposes a virtual reality (VR)-based fuzzy spatial relation knowledge extraction method for observer-centered vague location descriptions (VR-FSRKE). In VR-FSRKE, a VR scene is constructed, and users can interactively determine the fuzzy region corresponding to a location description under the simulated VR observer perspective. Then, a spatial region clustering mechanism is established to summarize the fuzzy regions identified by various individuals into fuzzy spatial relation knowledge. Experiments show that, on the basis of interactive scenes provided through VR, VR-FSRKE can efficiently extract spatial relation knowledge from many individuals and is not restricted by requirements of a certain place or time; furthermore, the knowledge obtained by VR-FSRKE is close to the knowledge obtained from a real scene.

Since the unsupervised segmentation of high-resolution remote sensing is a highly challenging task, the introduction of deep learning and processing may be a sensible choice to improve the quality of unsupervised segmentation. Unfortunately, any attempt to direct using unsupervised deep neural networks (UDNNs) to perform this task will be hindered by many obstacles: uncontrollable refinement processes, excessive fragmentation at the borders and excessive computing resource requirements. These obstacles can prevent us from obtaining acceptable results. To address this problem, this article proposes a hierarchical object-focused and grid-based deep unsupervised segmentation method for high-resolution remote sensing images (HOFG). Based on a grid approach, HOFG first adopt a lazy deep segmentation method (LDSM) to handle fragmentation and large image sizes. Then, a hierarchical and iterative segmentation strategy is introduced to reduce the accuracy expectation for the LDSM by means of a cascaded focus mechanism, making the entire segmentation process more controllable. HOFG can overcome all of the above obstacles while utilizing the high recognition ability of UDNNs. In experiments, HOFG are compared with shallow and deep unsupervised segmentation methods. The results show that HOFG can obtain fewer segments while maintaining a high accuracy. HOFG transform the unsupervised classification ability of UDNNs into a controllable and stable segmentation ability, making HOFG valuable for practical applications. The results show that on average, HOFG need only 81.73% as many segments as traditional shallow methods to achieve a high overall accuracy, and HOFG can obtain a 7.2% higher accuracy than a UDNN even when using only approximately 18% as many segments. HOFG can effectively and controllably utilize the recognition ability of UDNNs to achieve better unsupervised segmentation results.