An Optimal Faster-RCNN Algorithm for Intelligent Battlefield Target Recognition

With Its Calculation Method Based on the Natural Advantages of the Human Brain's Way of Thinking, Deep Learning Has a Great Advantage Over Traditional Methods for Image Recognition, Voice Recognition, and Text Processing. This Paper Uses Deep Neural Networks To Train Convolutional Neural Networks on Image Data Obtained From Unmanned Aerial Vehicles. in Addressing the Three Major Problems of the Original Model,i.e., Overfitting, Redundancy Recognition, and Low Recognition Accuracy, We Propose a Model Based on Characteristics of Images Acquired From Unmanned Aerial Vehicles, and Intelligently Generate a Set of Neural Networks that Can Quickly Recognize Multiple Types of Targets on the Battlefield. An Experiment Analyzes the Advantages and Disadvantages in Deep Learning Using Two Typical Models: Faster-Rcnn and Yolo v3. We Obtain Scenarios Applicable To Each Model, and Optimize and Improve the Recognition Accuracy of Faster-Rcnn. the Effectiveness of Our Algorithm Is Validated on a Large Experimental Dataset.