CODet: Component Object Detector Extracting Structural Features based on Target Characteristics

Deep learning technology has promoted the object detection task in the remote sensing field to move towards better performance and more demanding requirements. Except for rigid body objects, component objects with more complex characteristics remain a detection challenge. Its "partial rules and overall disorder" characteristic limits the model learning ability to the structural features. And the internal noise and relatively sparse arrangement are not conducive to optimizing the model by existing sample assignment strategies. We propose CODet to detect component objects in remote sensing scenes. It consists of a Cross-hierarchy feature Fusion Module (CFM) and a Noise-Sparse sample Assignment (NSA) strategy. CFM learns the potential representation and relative position relationship of components by fusing different level features. NSA redefines the optimization process of sample assignment. It aims to alleviate the problems of classification-localization misalignment and the positive-negative sample imbalance caused by the object's internal noise and sparse arrangement. The method is verified on the proposed COD dataset of six categories of component objects, reaching an average mAP/mAP50 of 54.3/86.0. To be closer to the task requirements of the practical remote sensing scene, we also propose a remote sensing large-scale images inference framework. It includes a dataset (APRoI, labeled with component objects and rigid body objects), a large-scale image inference strategy, and a set of evaluation metrics. With CODet as the core, the framework can effectively reduce inference time by 3 to 4 times on images with an average of more than 100 million pixels.