Enhancing Remote Sensing Object Detection with a Hybrid Densenet-LSTM Model

In both the military and civil fields, detection and identification of target objects in remote sensing imagery is utilized frequently. There are several models that have been developed for this purpose, however due to the influence of climatic conditions, barriers, and other factors, they are not always efficient at detecting target objects in remote sensing visuals. Challenges with small-target and unclear objects in photos, image clarity, and related issue detection and recognition of several targets. Therefore, it is crucial to find a solution for the challenge of reliably detecting target items in photographs. We present a remote sensing identification of objects method in this work that combines the power of LSTM and DenseNet. Our approach leverages the temporal dependencies captured by LSTM to improve sequential remote sensing data representation of features. Additionally, we utilize the deep convolutional architecture of DenseNet to extract rich spatial features from the imagery. By integrating these two techniques, we aim to increase object detection in remote sensing applications' resilience and accuracy. The effectiveness of our suggested method in achieving cutting-edge performance in terms of accuracy, precision, recall, F1-score, and loss is demonstrated by experimental findings on NWPU VHR-10 dataset compared with existing multi-scale CNN and YOLOV4.