Development Of Embedded Vision System Based On Jetson Nano

With the continuous progress of computer vision technology and electronic semiconductor technology,the computing power of various embedded devices has rapidly increased,and computer vision technology has also begun to be deployed on embedded devices.In underwater maintenance of water conservancy engineering facilities,the underwater environment is complex and there are various interference factors that restrict the widespread application of computer vision technology in practical engineering.This thesis studies the underwater 3D reconstruction technology based on structured light,and deploys it to a high-performance,lowpower,and portable embedded device to establish an embedded vision system based on Jetson Nano,achieving practical application in engineering.(1)Design and build the hardware platform of the embedded vision system based on Jetson Nano.Propose the functional requirements of the embedded vision system,design the hardware components of the system;considering the precision requirements of the system,taking into account factors such as computing performance and cost,select hardware devices such as the embedded core board,industrial camera,lens,line laser emitter,one-dimensional linear translation platform,etc.,and install and debug the hardware environment of the vision system.(2)Develop algorithms suitable for embedded underwater vision systems.The underwater structured light 3D reconstruction algorithm was implemented.To address the imaging defects commonly encountered in underwater images,an image enhancement algorithm was proposed to improve the image quality and improve the gray scale distribution of light stripe images.In addition,an improved laser stripe center point extraction algorithm was proposed based on the traditional grayscale centroid method.The algorithm was implemented in an embedded system while ensuring extraction accuracy.Finally,the camera imaging model,camera calibration,light plane calibration,and stitching vector calibration algorithms were optimized.(3)Design and implement the software of the embedded vision system.Install the operating system and configure the environment of Jetson Nano,install the development software for the visual system and the Qt interface design software;based on the functional requirements of the visual system,design the functional use cases,data exchange,and object design of the visual system software;use C++ programming to implement the visual system software,and introduce the functional settings of the embedded vision system software interface.(4)Conducted underwater 3D reconstruction experiments.Debug and run the vision system as a whole.Simulate the underwater environment in the laboratory,calibrate the parameters of the camera,light plane,and stitching vector,and conduct 3D reconstruction experiments on simulated underwater cracks and sea snail shells.The experimental results verified the feasibility of the visual system,and the reconstruction results have a certain accuracy.Finally,the factors affecting the reconstruction accuracy were analyzed,and shortcomings were found and improvement methods were proposed.