What is machine vision and what is the principle of machine vision?

Machine Vision (MV) is a technology and methodology that provides automated image-based detection and analysis for applications such as automated inspection, process control, and robot navigation, typically used in industry. Machine Vision is a term that encompasses a vast array of technologies, hardware and software products, integrated systems, actions, methods and expertise. Machine vision as a systematic engineering discipline differs from computer vision, a form of computer science. It attempts to integrate existing technologies in new ways and apply them to solve real-world problems. The term is commonly used in industrial automation environments, but is also used in other environments such as security and vehicle navigation.

The entire machine vision procedure generally involves three parts. First, form images.Then, automatically analyze images. Last, extract the required information.

The first step in the sequence of automated inspection operations is the acquisition of images, usually using cameras, lenses and lighting, which are also differentiated according to the subsequent processing. MV software packages and developed programs are developed based on them. Various digital image processing techniques are then used to extract the required information and decisions (e.g. pass/fail) are usually made based on the extracted information.

After acquiring an image, it needs to be processed. Multiple stages of processing are typically used in sequences that end with a desired result. Typical sequences may begin with tools such as filters to modify the image, followed by extraction of objects, followed by extraction of (e.g., measurement, code reading) data from those objects, followed by delivery of that data, or comparing it to a target value to create and deliver a “pass/fail” result. Machine vision image processing methods include.

Stitching/Registration: The combination of adjacent 2D or 3D images.
Filtering (e.g. morphological filtering)
Thresholding: Thresholding starts with setting or determining a gray value that will be used to separate parts of an image, and is sometimes used to simply convert parts of an image to black and white depending on whether the image’s gray value is below or above that gray value.
Pixel counting: counting the number of light or dark pixels
Segmentation:Splitting a digital image into multiple segments to simplify and/or change the presentation of the image to make it more meaningful and easier to analyze.
Edge Detection:Finding target edges
Color Analysis:Use color to identify parts, products, and objects, assess quality from color, and separate features using color.
Speckle Detection and Extraction:Detect discrete speckles of connected pixels in an image that serve as image landmarks (e.g., black holes in gray objects).
Neural network/deep learning/machine learning processing:Weighted and self-trained multivariate decision making. Around 2018, there was a major expansion in this area, using deep learning and machine learning to significantly extend machine vision capabilities.
Pattern recognition includes template matching. Finding, matching, and/or counting specific patterns. This can include the position of an object, which can be rotated, partially hidden by another object, or vary in size.
Barcode, Data Matrix and “2D Barcode” Reading
Optical Character Recognition (OCR): Automatic reading of text, e.g. serial numbers.
Measurement / Metrology: Measurement of object dimensions (e.g. in pixels, inches or millimeters)
Comparison with the target value (detection of defects) to determine a “pass/fail” result. For example, with code or barcode verification, the read value is compared to a stored target value. For measurements, the measured value is compared to the appropriate value and tolerance. To verify an alphanumeric code, the OCR’d (optical character recognition) value is compared to the appropriate value or target value. When checking for defects, the size of the measured defect is compared to the maximum value allowed by the quality standard.

Common outputs from automated inspection systems are pass/fail decisions. These decisions may in turn trigger mechanisms to reject failed items or issue alarms. Other common outputs include object position and orientation information from robot guidance systems. In addition, output types include digital measurement data, data read from codes and characters, counting and categorization of objects, displays of processes or results, stored images, alarms from automated spatial surveillance MV systems, and process control signals. This also includes user interfaces for the integration of multi-component systems and interfaces for automatic data exchange.

Advancevi company is committed to making breakthroughs in artificial intelligence deep learning technology and machine vision deep learning algorithms. After several years devotion and persistence, now it has introduced the latest machine vision equipment to inspect surface defects of various pipes, tubes, cables and wires. And through the machine vision, image processing algorithms, artificial intelligence, pattern recognition, ARM/FPGA/DSP embedded hardware and software development, the device aims to reduce the labor load, improve product quality, reduce costs and promote high-end manufacturing industry to achieve intelligent manufacturing.