Optimized Precision Computing Processor for Efficient XNOR-Based Target Recognition and Data Processing

Convolutional Neural Networks (CNNs) are highly effective in object detection, but their computational demands hinder real-time performance on mobile and resourceconstrained devices. This paper presents a novel object detection system that combines algorithmic and hardware optimization, achieving near-perfect detection with minimal computational overhead. The proposed approach leverages a binarized neural network (BNN) has been designed to maintain information stability despite deep quantization. A highly efficient FPGAbased object detection processor is introduced, capable of realtime processing with an accuracy of 99.9% Precision, Recall, and F1-Score. The processor efficiently handles variableprecision operations and eliminates the need for external DRAM access, significantly reducing hardware implementation costs and memory bandwidth. The proposed design outperforms previous FPGA-based methods by delivering exceptional throughput and accuracy, making it well-suited for integration into low-power mobile and embedded systems.