Abstract:Although many machine learning methods have been successfully applied for the object-based classification of high resolution (HR) remote sensing imagery, current methods are highly dependent on the spectral similarity between segmented objects and have disappointingly poor performance when dealing with different segmented objects that have similar spectra. To overcome this limitation, this study exploited a knowledge graph (KG) that preserved the spatial relationships between segmented objects and has a reasoning capability that can assist in improving the probability of correctly classifying different segmented objects with similar spectra. In addition, to assist the knowledge graph classifications, an image segmentation method generating segmented objects that closely resemble real ground objects in size was used, which improves the integrity of the object classification results. Therefore, a novel HR remote sensing image classification scheme is proposed that involves a knowledge graph and an optimal segmentation algorithm, which takes full advantage of object-based classification and knowledge inference. This method effectively addresses the problems of object classification integrity and misclassification of objects with the same spectrum. In the evaluation experiments, three QuickBird-2 images and over 15 different land cover classes were utilized. The results showed that the classification accuracy of the proposed method is high, with overall accuracies exceeding 0.85. These accuracies are higher than the K Nearest Neighbors (KNN), Support Vector Machine (SVM), Decision Tree (DT), and Random Forest (RF) methods. The evaluated results confirmed that the proposed method offers excellent performance in HR remote sensing image classification.

Learning by Asking Questions for Knowledge-Based Novel Object Recognition

Visual Question Generation for Class Acquisition of Unknown Objects

K-VQG: Knowledge-aware Visual Question Generation for Common-sense Acquisition

KGNet: Knowledge-Guided Networks for Category-Level 6D Object Pose and Size Estimation.

CON: Continual Object Navigation via Data-Free Inter-Agent Knowledge Transfer in Unseen and Unfamiliar Places

NOC-REK: Novel Object Captioning with Retrieved Vocabulary from External Knowledge

Sample-Efficient Learning of Novel Visual Concepts

Beyond Object Recognition: A New Benchmark towards Object Concept Learning

Novel Knowledge Graph- and Knowledge Reasoning-Based Classification Prototype for OBIA Using High Resolution Remote Sensing Imagery

YOLOOC: YOLO-based Open-Class Incremental Object Detection with Novel Class Discovery

Look Further to Recognize Better: Learning Shared Topics and Category-Specific Dictionaries for Open-Ended 3D Object Recognition

Learning Novel Objects Continually Through Curiosity

Class‐Relation Reasoning with Knowledge‐Transfer for Few‐Shot Object Detection

P-ODN: Prototype based Open Deep Network for Open Set Recognition

Enhancing Novel Object Detection via Cooperative Foundational Models

Knowledge-Experience Graph with Denoising Autoencoder for Zero-Shot Learning in Visual Cognitive Development.

Hybrid Knowledge Routed Modules for Large-scale Object Detection

Robo-CSK-Organizer: Commonsense Knowledge to Organize Detected Objects for Multipurpose Robots

Embodied Active Learning of Generative Sensor-Object Models

Training Self-localization Models for Unseen Unfamiliar Places via Teacher-to-Student Data-Free Knowledge Transfer

Breaking Immutable: Information-Coupled Prototype Elaboration for Few-Shot Object Detection