Salient Object Detection based Aircraft Detection for Optical Remote Sensing Images

Nowadays the use of Optical remote sensing images (RSIs) for detecting any particular object is being increased.Salient object detection is a fascinating aspect of optical RSIs (SOD). Regarding optical RSIs, there are a plethora of problems, like crowded backdrops, diverse object orientations, different object scales, etc. As a result, the execution of the current salient object detection models frequently suffers greatly. The relevance of information about edges, which is essential for producing correct saliency maps, is frequently overlooked by existing SOD models. To overcome this issue, this model uses Spatial Channel Attention U-Net (SCAU-Net) for detecting the edge maps. This model pop-out aircraft in optical RSIs using SOD. First, the input is sent into Encoder and SCAUNet simultaneously. The SCAU-Net provides salient edge cues and the encoders are used to give a good feature representation for salient objects. Then the output of the encoder is sent to the decoders. The decoders of the feature-merge module gives position attention to salient objects. The efficient edge map provided by SCAU-Net is used for improving the position attention cues. The final step is to combine all position attention cues to obtain the final output. The final output contains the aircraft detected in it. By seeing the obtained results we can say that our model can precisely and accurately detect the aircrafts present in the given input.