Cascade Object Detection and Remote Sensing Object Detection Method Based on Trainable Activation Function

Object detection is an important process in surveillance system to locate objects and it considered as major application computer vision. The Convolution Neural Network (CNN) based models have been developed by many researchers for object achieve higher performance. However, existing some limitations such overfitting problem lower efficiency small detection. remote sensing hasthe of low detecting the methods poor localization. Cascade Detection applied increase learning model. In this research, Additive Activation Function (AAF) a Faster Region CNN (RCNN) proposed AAF-Faster RCNN method has advantage better convergence clear bounding variance. Fourier Series Linear Combination activation function are used update loss function. Microsoft (MS) COCO datasets Pascal VOC 2007/2012 evaluate performance also analyzed benchmark dataset. analysis shows that model than state-of-art Pay Attention Them (PAT) To sensing, NWPU VHR-10 data set test method. mean Average Precision (mAP) 83.1% PAT-SSD512 81.7%mAP 2007