CN211906310U - Zynq-based machine vision detection system - Google Patents

Abstract

The utility model discloses a machine vision detecting system based on Zynq. The utility model comprises an FPGA core and an ARM core; the FPGA core comprises a frame trigger module for signal filtering, a camera driving module, an FPGA processing acceleration module for image processing, a display driving module and a VDMA module; the ARM core comprises an AXI bus, a storage module, an ARM and a USB HOST module, wherein the ARM and the USB HOST module are used for running a Linux system, the AXI bus is connected with the VDMA module, the storage module and the ARM, the ARM is connected with the USB HOST module, the USB HOST module is used for connecting a keyboard and a mouse, and the ARM is used for connecting a network. The utility model discloses machine vision detecting system's real-time performance can be improved greatly to can carry out the secondary development.

Description

Zynq-based machine vision detection system

Technical Field

The utility model relates to a machine vision detects technical field, specific machine vision detecting system based on Zynq that says so.

Background

Machine vision is to use machine to replace human eyes to identify, judge and measure target objects, and mainly researches on simulating human vision function by using a computer. The machine vision technology is a comprehensive technical concept and comprises a vision sensor technology, a light source illumination technology, an optical imaging technology, a digital image processing technology, an analog and digital video technology, a computer software and hardware technology, an automatic control technology and the like. Machine vision is not only characterized by simulating human eye functions, but more importantly, it can accomplish some tasks that human eyes cannot.

The machine vision detection system is an image processing system for automatically detecting and analyzing products produced in industry in the industrial automatic production process. For different industrial products, the functions of the required intelligent camera system are slightly different, and can be roughly divided into functions of defect detection, size measurement, color identification, feature extraction, feature positioning, bar code, two-dimensional code identification and the like.

In the industrial production process, compared with the traditional inspection method, the machine vision technology has the greatest advantages of rapidness, accuracy, reliability and intellectualization, and has irreplaceable effects on improving the consistency of product inspection and the safety of product production, reducing the labor intensity of workers and realizing efficient safe production and automatic management of enterprises.

At present, machine vision systems in the industrial field are expensive, rely on import mostly, and are bulky, and the machine vision systems have single function for general vision processing and have slow speed of image acquisition and processing. And the maintenance cost in the use process is high, and secondary development cannot be carried out. Most of the image data processing using PCs have difficulty reaching the requirement of real-time processing. And has the defects of large power consumption and difficult maintenance. Moreover, the machine vision system is large in size, and a PC is mainly used as an image processing platform, so that the size of the whole vision detection system is inevitably increased.

With the development of science and technology, the requirements of various industrial fields on a machine vision system are different from those of the previous days, and with the higher integration degree, the shorter research and development period is required; with the improvement of the precision degree, the volume of a vision system is required to be smaller and smaller; with the increase of production speed, the real-time processing speed of the vision system is required to be higher and higher. In addition, the types of the industrial fields in China are complete, the requirement on a machine vision system is high, and the difference of the machine vision systems in different industries mainly lies in an image post-processing method. At present, no universal machine vision system capable of being developed secondarily exists.

The existing machine vision detection system has poor real-time performance mainly because of low acquisition and processing speed, which causes low production efficiency. At present, most machine vision systems use chips such as a DSP, a CPU and the like to complete hardware design of the system, and do not adopt a Zynq platform. The Zynq platform comprises FPGA resources and an efficient ARM core processor, so that the speed of image acquisition, image processing and image transmission of the machine vision detection system is greatly improved, and the effect of real-time processing can be achieved.

SUMMERY OF THE UTILITY MODEL

To the above-mentioned weak point that exists among the prior art, the to-be-solved technical problem of the utility model is to provide a machine vision detecting system based on Zynq.

The utility model discloses a realize that the technical scheme that above-mentioned purpose adopted is: a Zynq-based machine vision detection system comprises an FPGA core and an ARM core;

the FPGA core comprises a frame trigger module for signal filtering, a camera driving module, an FPGA processing acceleration module for image processing, a display driving module, a first VDMA module, a second VDMA module and a third VDMA module for data caching, wherein the frame trigger module receives an external level signal, the camera driving module is connected with an external camera, the frame trigger module is connected with the camera driving module, the camera driving module is connected with the first VDMA module, the FPGA processing acceleration module is connected with the second VDMA module, the display driving module is connected with the third VDMA module, and the display driving module is used for connecting a display;

the ARM core comprises an AXI bus, a storage module, an ARM and a USB HOST module, wherein the ARM and the USB HOST module are used for running a Linux system, the AXI bus is connected with the first VDMA module, the second VDMA module, the third VDMA module, the storage module and the ARM, the ARM is connected with the USB HOST module, the USB HOST module is used for connecting a keyboard and a mouse, and the ARM is used for connecting a network.

The memory module is DDR 3.

The VDMA module adopts an IP core of Xilinx company.

The utility model has the advantages of it is following and beneficial effect:

1. the utility model utilizes FPGA technology and embedded Linux technology, firstly uses the FPGA kernel of PL part to drive the image acquisition sensor, carries out image acquisition and input, and then stores in DDR 3; the FPGA is triggered by logic in parallel, so that the processing speed is extremely high, and the real-time performance of a machine vision detection system can be greatly improved; an embedded Linux operating system is implanted on an ARM core of the PS part, and further processing, outputting and displaying are carried out on image data collected and processed by the PL part.

2. The utility model discloses the volume is very little, has QT software development environment and OpenCV storehouse in the Linux of embedded moreover for user's secondary development has shortened the cycle of research and development greatly, reduces the cost of research and development.

Drawings

FIG. 1 is an overall structure diagram of the present invention;

fig. 2 is a functional schematic diagram of a frame trigger module and a camera driving module according to an embodiment of the present invention;

fig. 3 is a block diagram of the remote network upgrade of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only some embodiments of the present invention, not all embodiments.

Zynq refers to a chip produced by Xilinx corporation, the chip is heterogeneous in multi-core, data analysis and hardware acceleration can be realized, and functions of a CPU, a DSP, an ASSP and mixed signals are highly integrated on a single device. The chip is generally divided into a PL part and a PS part, wherein the PL part refers to an FPGA core, and the PS part refers to an ARM core. The time sequence logic design can be realized on PL through a programmable hardware description language Verilog language, and a cuttable Linux system can be embedded in PS part.

The FPGA core of the PL part is used for driving an image acquisition sensor to acquire and input images, and then the images are stored in a DDR3 (storage module). The FPGA is triggered by logic in parallel, so the processing speed is very high, and the real-time performance of the machine vision detection system can be greatly improved. The image data collected and processed by the PL part is further processed and output and displayed on an ARM core of the PS part, and secondary development of image processing can be performed.

The machine vision detection system provided by the invention is divided into a USB HOST module, a camera driving module, an FPGA processing acceleration module, a VDMA module, a frame trigger module, an embedded Linux system module, a network module, a display driving module and the like, and is shown in figure 1. The MT9V034 camera generates images through a camera driving module, and then the images are stored into a DDR3 cache through a VDMA module, wherein the three VDMA modules are mainly used for caching data, and specifically comprise: the camera driving module writes data into the DDR3 through the first VDMA module, the FPGA processing acceleration module reads and writes data with the DDR3 through the second VDMA module, and the display driving module reads data from the DDR3 through the third VDMA module. At this time, the embedded Linux system module can read image information from the DDR3, then perform a series of image processing operations, a specific image processing algorithm is determined according to a usage scene and a detection object, then transmit a processed image interface to the display driver module through the VDMA module, then the display driver module drives an external display screen to display, a user can operate QT software in the Linux system through a keyboard and a mouse to adjust some parameters, and some necessary data can be transmitted to the server through the network module.

In the design of a sequential logic circuit of a PL part of a Zynq platform, main functional modules are divided into a camera driving module, a frame triggering module, an FPGA processing acceleration module, a display driving module and a VDMA module. The camera driving module is mainly used for completing image data receiving and register configuration tasks of the MT9V034 camera, converting an optical image into a digital image and configuring parameters such as resolution, exposure, shutter time and gain. As shown in figure 2, the filtering of external level signals is mainly completed by the frame trigger module circuit, the external level signals can be used for detecting the materials by aiming at the problem of inconsistent arrival time of the materials on the flow production line, when the infrared photoelectric switch is shielded by the materials, the level signals can be returned, the trigger level is generated by processing the frame trigger module, and the trigger level is transmitted to the camera driving module to take a picture for one time.

The FPGA processing acceleration module mainly completes algorithms for image processing, such as sobel filtering, OSTU, binarization and the like, and the algorithm is realized by using hardware, so that the processing speed is high, and the effect of real-time processing can be achieved. In industrial production, efficiency is an important element, and the faster the material detection on a production line, the better, so that an algorithm for completing image processing by using an FPGA is advantageous.

The logic of the display driving module mainly simulates the data format of the HDMI, and the display driving module has the advantages of reducing the cost and saving the area of a printed circuit board. The VDMA module uses an IP core provided by Xilinx, and the IP functions to perform read-write operations on the DDR3 to complete data access.

An embedded Linux operating system is transplanted and cut in a PS part of the Zynq platform, and U-boot source codes and kernel source code packages which are adaptive to Zynq chips are used. The Linux operating system provides a uniform API (application programming interface) interface for the driving program, so that the interface functions can be conveniently called in a system task, and the transmission of data from the acquisition to the system is completed. The image collected by the camera driving module is mainly used for enabling the system to configure and operate the image information collected by the camera. The display driving module is mainly used for frame buffer of the display, and real-time refreshing is carried out on virtual addresses in the system, so that the display can only see the change of the picture. In addition, the development of the interface program of the embedded Linux system is realized.

To the detection target of difference, machine vision system need realize different image processing algorithm, the utility model provides a machine vision system can carry out quick secondary development, has also made things convenient for later maintenance and upgrading when having improved the commonality. Aiming at different target detection, when secondary development is carried out, only the algorithm in the application program needs to be modified, and the OpenCV library is transplanted and a plurality of image processing algorithms are packaged in the OpenCV library, so that the development time is saved. The later upgrade and maintenance of the system can be realized through a remote network, and the specific implementation manner is as shown in fig. 3, the generated system firmware is transmitted to an eMMC (onboard SD card) on a board through an ethernet in a UDP manner, and when the system is powered on again, the system first executes the firmware FLASH refresh in the starting process, and then completes the upgrade of the system.

The Zynq-based machine vision detection system provided by the invention has the following functions: acquiring images in real time, and transmitting and processing the images; libraries for visual processing such as QT and OpenCV can be transplanted on an embedded Linux operating system, so that secondary development is facilitated; the level trigger signal can be adopted for collecting images, mainly aiming at the problem of inconsistent arrival time of materials on a flow production line, an infrared photoelectric switch can be used for detecting the materials, and when the infrared photoelectric switch is shielded by the materials, the level signal can be returned, so that a machine vision system is triggered to take a picture; the display part supports VGA and HDMI display output, and an image interface can be displayed on a display with a corresponding interface directly; in the input part, the input of a keyboard and a mouse is supported; in the networking part, the system can be used as an independent module to exchange information with the server; in addition, the embedded Linux operating system of the system can be remotely upgraded through a network, so that the overall adaptability of the system is greatly improved, the cost of research and development design again is saved, and the efficiency of industrial production is improved.

The Zynq-based machine vision detection system provided by the invention comprehensively utilizes the strong parallel computing capability of the FPGA, so that the image processing can achieve the real-time purpose, and the efficiency in industrial production is improved. The QT interface design of human-computer interaction is completed by utilizing the real-time multitask characteristic of the embedded Linux operating system, and the output display of the HDMI is carried out. The power consumption is reduced, the application range of the system is improved, the image processing function is small in size and high in performance, and meanwhile later-stage maintenance and upgrading are easier to perform.

The utility model is suitable for a following each big trade:

1. hardware plastic industry

Hardware is various in kind, and the shape is various, and is not of uniform size, and the work of artifical detection and letter sorting is very complicated and makes mistakes easily, not only causes very big waste and makes enterprise reputation receive the harm easily for the enterprise, and machine vision detecting system not only can do zero error rate and practiced thrift the cost of labor, has played irreplaceable effect to improving product quality.

2. Semiconductor industry

The semiconductor and electronic industry is an industry requiring high precision, high efficiency and zero error, and due to the rapid development of modern processing technology, the requirement on detection is higher and higher, and the problem is just solved by machine vision. For the detection of the welding of the printed circuit board, the existence of the device with welding missing can be detected.

3. Packaging industry

In modern automated production, various inspection, measurement and part identification applications are involved, such as integrity checks in the packaging industry, beverage filling level checks, print quality checks, etc.; the common characteristics of the products are that the mass production and the requirement on appearance quality are very high, the manual detection is completed before, the labor cost and the management cost are increased, and the inspection qualified rate of 100 percent still cannot be ensured. Machine vision is used to replace human eyes for measurement and judgment.

4. Pharmaceutical industry

In the medicine production industry, for the detection of the tablet package, the medicine package on the conveying belt is detected, and whether the defects of mixed loading, neglected loading, medicine breakage and the like exist is judged. And detecting the label of the medicine bottle, reading the printed characters of the label, judging the printing quality and identifying the type.

5. Bearing processing industry

Detect the defect of bearing ball, when having ball disappearance or ball damaged in the bearing, the software can detect immediately, can send alarm signal simultaneously. The camera is fixed above the rotating platform, and the camera acquires images after the bearing rotates to a position right below the camera and simultaneously detects the images. The detection method is based on template comparison, the template is created by a standard product, and when the acquired image cannot be successfully compared with the template, the product is defective.

6. Automobile hub detection

The hub is classified, and the hub with various models on the production line is mixed and needs to be classified according to different models. A mesa array camera is installed on the assembly line, and image acquisition is carried out to each wheel hub, and the software is according to the different automatic identification wheel hub models of wheel hub aperture size to give the manipulator with model information transfer, the manipulator snatchs wheel hub according to the model of difference and classifies.

The above, only be the concrete implementation of the preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto, and any person skilled in the art is in the technical scope of the present invention, according to the technical solution of the present invention and the utility model, the concept of which is equivalent to replace or change, should be covered within the protection scope of the present invention.

Claims (3)

1. A Zynq-based machine vision detection system is characterized by comprising an FPGA core and an ARM core;

the FPGA core comprises a frame trigger module for signal filtering, a camera driving module, an FPGA processing acceleration module for image processing, a display driving module, a first VDMA module, a second VDMA module and a third VDMA module for data caching, wherein the frame trigger module receives an external level signal, the camera driving module is connected with an external camera, the frame trigger module is connected with the camera driving module, the camera driving module is connected with the first VDMA module, the FPGA processing acceleration module is connected with the second VDMA module, the display driving module is connected with the third VDMA module, and the display driving module is used for connecting a display;

the ARM core comprises an AXI bus, a storage module, an ARM and a USB HOST module, wherein the ARM and the USB HOST module are used for running a Linux system, the AXI bus is connected with the first VDMA module, the second VDMA module, the third VDMA module, the storage module and the ARM, the ARM is connected with the USB HOST module, the USB HOST module is used for connecting a keyboard and a mouse, and the ARM is used for connecting a network.

2. The Zynq-based machine vision inspection system in accordance with claim 1, wherein the memory module is DDR 3.

3. The Zynq-based machine vision inspection system in accordance with claim 1, wherein the VDMA module employs an IP core of Xilinx corporation.

Images