Research on Remote Sensing Image Detection Algorithm Based on Improved Mask RCNN
With the continuous development of the field of aerial remote sensing,the detection of small targets in this scenario has become an important work in the current research field.Based on aerial remote sensing images,this paper proposes an optimiza-tion method for small target detection in the field of aerial remote sensing.In order to improve the practicability and accuracy of the algorithm in small target detection,this paper adds a spatial attention mechanism module to the Mask RCNN algorithm to reduce the noise of the image background,and uses CIOU as the bounding box regression loss function to optimize.And then it uses the Kmeans clustering algorithm instead of the original algorithm to generate a detection anchor box that more matches the small target.The improved Mask RCNN has a detection accuracy of 61.89mAP under the aerial remote sensing image data set.Compared with the current mainstream remote sensing image detection algorithm R-FCN,the detection accuracy has increased by 17%,and compared with Mask RCNN by 2.4%,reaching the best detection effect under the current conditions.
国家科技期刊平台
NSTL
万方数据
