Research And Implementation Of Target Tracking System Based On Edge Intelligence

Techniques such as face detection,face recognition is applied target tracking system based on edge intelligence to edge device,it lets users can efficiently screen out the moving tracks of target people on edge device which need enhance requirement on the performance of edge device 。 Due to its outstanding classification ability,Deep learning algorithms become the preferred method for face detection and recognition。It is very important of research emphasis that target tracking system how to deploy convolutional neural networks on Edge device。This thesis has conducted research on this issue,and the main work is as follows:1)This thesis proposes a lightweight face detection algorithm to address the issue of convolutional neural networks having too much computational complexity and difficulty deploying on embedded devices.This algorithm takes YOLOv5 s as the benchmark model,replaces its backbone network with a lightweight network Mobile Net,and introduces attention mechanism CBAM and model pruning scheme.The performance testing of the proposed improved algorithm shows that it greatly reduces the model volume while maintaining high accuracy,achieving lightweight face detection.2)In response to the high concurrency data transmission requirements of multiple nodes,this thesis ported the Sparrow RV processor core,bus system,and interrupt system based on RISC-V architecture.This processor uses the RV32 I basic integer instruction set,M,and Zicsr extensions,uses the ICB bus protocol and PLIC platform level interrupt controller,and has a two-level pipeline.After Coremark run point testing,this design has low power consumption and sufficient performance to meet the requirements of application scenarios.3)Design a target tracking system,with the hardware system divided into a main control node and a monitoring node.The main control node consists of a microprocessor module,communication module,etc.,responsible for collecting and processing data;The monitoring node consists of a microprocessor module,communication module,video acquisition module,etc.It is responsible for video acquisition,storage,wireless transmission,and has the function of facial detection.By selectively retaining the collected information,meaningless information is reduced.The test results indicate that the target tracking system designed in this thesis works reliably and basically meets the requirements of practical applications.