Robot Vision: A Comprehensive Analysis of Principles, Components, and Applications
In today's era of rapid technological development, robot vision technology is gradually becoming one of the key technologies in the field of automation. According to data, the global market size of mechanical vision reached $ 11.4 billion in 2021, and by 2022, this number is expected to grow to $ 12 billion, indicating a continuous upward trend. This indicates that robot vision technology is receiving increasing attention and application worldwide.
1, Visual project
Basic functions
Recognition: The recognition function mainly involves identifying the characteristics of the target object, such as its appearance. Among them, the accuracy and speed of barcode recognition are important indicators for measuring recognition ability.
Measurement: The measurement function can obtain the unit of size of the image and accurately calculate the geometric dimensions of the target object in the image. High precision and complex shape measurement are the advantages of machine vision in this function.
Localization: Localization is currently a widely used field that can obtain two-dimensional and three-dimensional position information of targets, with accuracy and speed being the main measurement indicators.
Detection: The detection field accounts for 50% of machine vision functions, and implementing calculations is challenging, primarily involving post-assembly appearance detection and appearance scratch defect detection.
Application scenarios
Machine vision combined with industrial robots is mainly used to guide robot movement. The specific scenarios can be divided into grasping, detection, and processing etc. The grabbing category can be subdivided into applications such as loading and unloading, palletizing, sorting, etc The process category mainly includes application scenarios such as gluing, polishing, welding, etc., and is mainly focused on grasping.
2, Composition and principles of the visual system
a. System Composition
Visual camera: Its main function is to capture images and collect image information.
Light source: Provide a stable light source scene for the visual system, so that the robot can obtain clearer images.
Computer hardware: including CPU, memory, hard disk, etc., mainly responsible for processing images, algorithm calculations, and storage.
Robots: Receive visual data, obtain physical coordinates, and execute automated production tasks based on visual instructions.
Mechanical device: including fixtures, conveyor belts, elevating seats, and other peripherals, the main function is to assist the robot in completing physical operations.
b. System classification
Monocular vision: This is a commonly used visual system that uses a single industrial camera for image acquisition, typically only able to capture two-dimensional images, and is widely used in the field of intelligent robots. However, due to issues with image accuracy and data stability, it is often necessary to work together with other types of sensors.
Binocular vision: consisting of two cameras, it uses the principle of triangulation to obtain depth information of the scene, and can reconstruct the three-dimensional shape and position of surrounding objects. The principle is similar to that of the human eye and is relatively simple.
Multi vision: By using multiple cameras, blind spots can be reduced, and the probability of erroneous detection can be lowered. It is widely used in the assembly field of industrial robots, and can accurately identify and locate the measured object, improving the intelligence and positioning accuracy of assembly robots.
c. Imaging principle
Visual imaging mainly converts the detected object into an image signal based on image acquisition devices (CMOS and CCD) and transmits it to a dedicated image processing system. Convert the brightness and color information of pixel distribution into digital signals. The image processing system extracts features of the target based on these signals, such as area, quantity, position, length, etc., and outputs results according to preset tolerance and other conditions, including size, angle, number, qualified/unqualified, presence/absence, etc., in order to achieve automatic recognition function, and then control the action of on-site equipment based on the discrimination results.
3. The difference between CCD and CMOS
CCD cameras use CCD to convert optical images into digital signals for transmission. CCD image processing sensors use one or a few output nodes for signal readout, with good transmission consistency and the ability to read the entire image information. However, the bandwidth of the output signal needs to be amplified, resulting in high power consumption.
CMOS cameras use CMOS to convert optical images into digital signals for transmission, using a single pixel for transmission, which can achieve single pixel signal amplification and extremely high image scanning rate, but there are defects in signal consistency.
The application of robot vision technology in the field of automation is constantly expanding and deepening. From various aspects such as the growth of its market size, the diversity of functions, the complexity of system composition, and the scientific nature of imaging principles, this technology will undoubtedly play a more important role in many fields such as industrial production and intelligent robots, in the future.