Analysis of machine vision color imaging applications

Introduction

Machine vision has been widely used in various industries around the world, such as semiconductors, electronics, automobiles, food, medicine, textiles, and transportation. In China, with the improvement of supporting infrastructure, the accumulation of technology and funds, various industries have widely adopted image and machine vision technology for industrial automation and intelligentization, and have gradually begun to apply it in industrial sites, such as drug testing and packaging in the pharmaceutical industry, printing color testing in the printing industry, surface quality testing of PCB boards, quality testing of LCD screens, etc. It can be said that machine vision has penetrated into all aspects of our lives, production, and work. However, with the increasing application, the acquisition of grayscale images can no longer fully meet the needs of applications, and color image acquisition technology is particularly urgent and important. At present, color image acquisition applications have been successfully introduced into the new generation of LCD display industry and PCB inspection industry. So, how to achieve the leap from grayscale image detection to color image detection?

Screenshot of the EasyColor Microscopy application content example in Open eVision Studio

Classification of color image components

The three primary colors of color images, RGB (red, green, and blue), are generally suitable for color reproduction and are often used on monitors and cameras, while the color system based on "saturation/chroma/hue" is used in machine vision applications. Therefore, color image processing and analysis software must be able to distinguish the colorless (
grayscale in order to interpret the color more intuitively and completely, which is very effective for detecting and segmenting color objects and removing light effects. Displaying the hue component (H component in the HIS color space) does not change the object coloring, shadows, and brightness intensity. Similarly, this hue component has its differences in intrinsic color or physical objects.

Therefore, when performing color image processing or analysis, it is necessary to first convert the RGB image to another color space, such as LSH, ISH, or YSH.

Screen shot of the EasyColor food inspection application example in Open eVision Studio

Color Image Conversion

Converting colors is a time-consuming task. To solve this problem, color image analysis software uses high-resolution color lookup tables; a lookup table is an array of values ​​that contains the correspondence between inputs and outputs. Once the lookup table is pre-calculated, it can be checked for the transformation applied to each image.

The values ​​in the lookup table can be used to convert images from one color system to another, and can also be set to apply user-defined transformations such as gamma compensation and color balance correction. In addition, some operations will "check" the contents of the lookup table to avoid the need to store the converted image. For example, the H component of the image's ISH color space can be changed while the image is still stored in RGB format. [page]

Color Segmentation of Color Images

In addition to traditional color image processing functions such as Bayer format to color space conversion or color balance correction, color image analysis software also provides powerful color analysis functions that allow users to classify color objects and detect defects.

"Thresholding" is the basic mechanism for segmentation, which can distinguish objects from the image background based on contrast. In the color image part, you can also freely specify the value range of specific objects. The user can select the value range of each component, for example, select "olive green" pixels based on hue only, plus a loose discrimination of saturation and chroma to eliminate surface lighting effects.

Another more effective way to distinguish different objects in the scene is to consider a set of specific colors and specify each pixel that is closest to these colors. The color image segmentation function allows you to specify the "level" (integer index value) of each pixel to decompose the color image into different areas.

Application extension of color image processing component library

The above EasyColor color image processing function is usually used in conjunction with other image processing component libraries. For example, the color image thresholding technology can be used with the EasyOCR, EasyMatrixCode or EasyBarCode component libraries to read characters, 2D matrix codes or barcodes on color printing or color objects respectively. The color image segmentation technology can be used with the EasyObject component library of Open eVision to perform blob analysis on the segmented area. In addition, color image processing can also be integrated into other software modules of Open eVision, such as Open eVision's component library EasyMatch specifically for grayscale matching.

Screenshot of the EasyColor Pharmaceutical Inspection application example in Open eVision Studio

Other color image conversion functions

Other color image conversion functions often include color balance (also known as white balance), which adjusts the intensity of colors (usually the three primary colors of red, green, and blue) across the entire domain. The main goal of the adjustment is to correctly present specific colors - especially neutral colors. In machine vision applications, color imbalance is mainly caused by the type of light used, and it may also be due to errors in the calibration of the image acquisition device. At the

same time, the best Bayer format to RGB conversion function, pseudo coloring function (enhanced grayscale image effect) and component expansion function are also required for color image processing.

Application Examples

Color image processing is most commonly used in the following industries, including: fruit and vegetable or packaged food inspection in the food processing industry, drug inspection in the pharmaceutical industry, FPD inspection (flat panel display), PCB inspection (printed circuit board) in the electronics industry, and printed product inspection in the printing industry.

Latest Trends

The latest market trend for color image acquisition applications is: color conversion through hardware.

For example: Euresys' Grablink Quickpack ColorScan is compatible with Camera Link line scan cameras and uses the FPGA on the motherboard to provide accelerated image pre-processing for color scanning applications. These functions include: scanning delay compensation (by acquiring RGB images multiple times, reconstructing the true RGB image according to the speed of the object's movement, thereby eliminating the problem of color misalignment), light field unevenness correction (compensating for uneven lighting in the image), lookup table and white balance correction, thereby greatly simplifying and accelerating color document scanning or PCB inspection.

Another example is the new Grablink Quickpack CFA, which is compatible with Camera Link Bayer and RGB area scan cameras. The card has pre-processing functions for Bayer decoding, white balance, brightness extractor, and four LUT operations. (end)