For remote sensing object detection, fusing the optimal feature information automatically and overcoming the sensitivity to adapt multi-scale objects remains a significant challenge for the existing convolutional neural networks. Given this, we develop a convolutional network model with an adaptive attention fusion mechanism (AAFM). The model is proposed based on the backbone network of EfficientDet. Firstly, according to the characteristics of object distribution in datasets, the stitcher is applied to make one image containing objects of various scales. Such a process can effectively balance the proportion of multi-scale objects and handle the scale-variable properties. In addition, inspired by channel attention, a spatial attention model is also introduced in the construction of the adaptive attention fusion mechanism. In this mechanism, the semantic information of the different feature maps is obtained via convolution and different pooling operations. Then, the parallel spatial and channel attention are fused in the optimal proportions by the fusion factors to get the further representative feature information. Finally, the Complete Intersection over Union (CIoU) loss is used to make the bounding box better cover the ground truth. The experimental results of the optical image dataset DIOR demonstrate that, compared with state-of-the-art detectors such as the Single Shot multibox Detector (SSD), You Only Look Once (YOLO) v4, and EfficientDet, the proposed module improves accuracy and has stronger robustness.

No abstract

Aiming at the blind angle in detecting weak signals of the same frequency by Duffing oscillator, a novel method of dephasing for the driving signals is proposed to eliminate the blind angle. According to the characteristic of weak signals, expression of blind angle is analyzed, and then the range of blind angle is found out, which corresponds to the amplitude of a new driven signal synthesized from the original driven signals and the unknown one. By making the original driven signal phase shift a degree of π, detection for the same frequency signal can be realized when the synthesized signal is located in the blind angle region, whose feasibility is proven by an experiment that it remains in chaotic status in the case of blind angle but becomes a great period status after the original driven signal's phase is dephased by π. To overcome the drawbacks of qualitative analysis and distinguish effectively different status in signal detection course, a detection statistics based on likelihood-Halmiton system is constructed. On the basis of it, a diagram of detection for any frequency signal is drawn. The key point is to make it as an unknown signal's frequency range where there are two adjacent frequency values whose corresponding detection statistics both located in the range of intermittent chaotic status, while one of them is just corresponding to the maximum value of the detection statistics. By simulations of different circumstances, detection statistics for numerical ranges of chaos, intermittent chaos, and great period is summarized. Furthermore, detection for any frequency signal may be realized by use of the numerical range. It is shown that the proposed method could not only provide quantitative judgment for the system status, but improve the signal detection performance. Also, it could be combined well with the traditional oscillator array method or adaptive step intermittent chaotic oscillator method, which further can improve some existing signal detection methods with Duffing oscillator.

No abstract

No abstract