Deep learning technology has been extensively explored by existing methods to improve the performance of target detection in remote sensing images, due to its powerful feature extraction and representation abilities. However, these methods usually focus on the interior features of the target, but ignore the exterior semantic information around the target, especially the object-level relationship. Consequently, these methods fail to detect and recognize targets in the complex background where multiple objects crowd together. To handle this problem, a diversified context information fusion framework based on convolutional neural network (DCIFF-CNN) is proposed in this paper, which employs the structured object-level relationship to improve the target detection and recognition in complex backgrounds. The DCIFF-CNN is composed of two successive sub-networks, i.e., a multi-scale local context region proposal network (MLC-RPN) and an object-level relationship context target detection network (ORC-TDN). The MLC-RPN relies on the fine-grained details of objects to generate candidate regions in the remote sensing image. Then, the ORC-TDN utilizes the spatial context information of objects to detect and recognize targets by integrating an attentional message integrated module (AMIM) and an object relational structured graph (ORSG). The AMIM is integrated into the feed-forward CNN to highlight the useful object-level context information, while the ORSG builds the relations between a set of objects by processing their appearance features and geometric features. Finally, the target detection method based on DCIFF-CNN effectively represents the interior and exterior information of the target by exploiting both the multiscale local context information and the object-level relationships. Extensive experiments are conducted, and experimental results demonstrate that the proposed DCIFF-CNN method improves the target detection and recognition accuracy in complex backgrounds, showing superiority to other state-of-the-art methods.

No abstract

No abstract

No abstract

Object detection in high-resolution remote sensing images has been attracted increasing attention in recent years owing to the successful applications of civil and military. However, there are many critical challenges deciding the performance of object detection in large-scale complex remote sensing image. One of these challenges is how extract and enhance the discriminative features without the top-down feedback mechanism for the existing convolutional neural network (CNN). To cope with this problem, a novel object detection algorithm based on direct feedback control for CNN (DFCCNN) is proposed in this paper. The DFCCNN combines a region proposal network with an object detection network to generate the proposals and to detect the object separately. Initially, a candidate region proposal network (CRPN) is implemented to extract candidate regions within the remote sensing image. Then multi-class objects detection feedback network (MODFN) propose a new top-down feedback mechanism based on the traditional feedforward network to detect the objects. A direct feedback loop (DFL) and a feedback control layer (FCL) are contained in the feedback network. The DFL propagates the posterior information directly from the top layer without depending on the rest of the network and the FCL make full use of top-down information to inhibit object-irrelevant neurons and emphasize object-relevant neurons. Through the addition of direct feedback control mechanism, these object-relevant neurons can be emphasized by taking feedback information of top layer into feature extraction, whereas these object-irrelevant neurons can be inhibited effectively by pruning the neural pathway. The proposed DFCCNN model is able to extract more discriminative low-level features under the guidance of the high-level information. Some experiments on NWPU VHR-10 data set and aircraft data set are induced, and the experimental results show that the proposed method can achieve a higher accuracy of object detection in remote sensing image with various complex background clutter.INDEX TERMS Object detection, convolutional neural network, direct feedback loop, feedback control layer, remote sensing image.