Research And Design Of Embedded 3D Industrial Camera Control System Based On Binocular Vision

When traditional two-dimensional industrial cameras detect complex parts,they do not distinguish the flatness,angle,depth,attitude and other 3d information of the measured workpiece well.People often can only set up multiple cameras to detected those workpieces,which may cost more and complicate the detect system.As one of the most widely used methods in 3D industrial cameras,binocular vision technology has the advantages of simple modeling,excellent effect and low cost.It is realized by using binocular camera to imitate the principle of binocular imaging,and the image acquired by binocular camera usually needs to be processed in the camera controller.Most of the existing 3D industrial camera use software to control the CPU for image processing.This kind of operation on the PC has the disadvantages of low system integration and low portability;In embedded systems such as X86 and arm-based systems,the operation time is long and the effect is poor;In optimized systems based on DSP and GPU,the circuit board's cost is high and the market demands is low.In this thesis,a low-cost ZYNQ series chip based on ARM+FPGA structure is established as the control system CPU scheme.This scheme can not only complete the serial scheduling of related tasks,improve the system integration,but also make use of the excellent parallel computing capability of FPGA to complete the conventional image processing computing tasks in binocular vision,and improve the system running speed.This thesis mainly studies and designs the following aspects:(1)Analyzing the principle of binocular vision 3D imaging,introducing the gaussian filtering algorithm,histogram equalization algorithm and extended Sobel operator based edge detection algorithm commonly used in embedded image processing,introducing and improving the Census transform-based stereo matching algorithm.(2)Designing he hardware modules involved in the embedded 3D industrial camera,which includes CMOS image sensor,HDMI image display,internal transmission scheme based on DMA and external transmission scheme based on Ethernet.A custom IP core is designed based on Verilog language to drive the image sensor and HDMI interface.(3)The detailed design and optimization of the image processing algorithm proposed above on FPGA hardware are made.According to the parallel operation characteristics of FPGA,the design greatly reduces the algorithm implementation cycle,and at the same time,it optimizes the utilization rate of hardware resources by combining with the targeted optimization method.The system test results show that the image processing results of the embedded camera control scheme designed in this thesis are similar to those achieved by the software on PC,and the computation time is greatly reduced.The system can run embedded QT and 3D graphics library compatible,convenient for subsequent application migration.The resource consumption in the controller chip is appropriate,and the overall performance parameters of the system meet the design objectives,so as to achieve the performance indexes of 3D industrial cameras on the market.