Juhi Ranjan, Rongali Mahesh, Shivam Chaubey
PhD Research Scholar, ICAR-IARI, New Delhi

Machine vision, an integral facet of artificial intelligence and computer vision, has revolutionized industries by enabling machines to interpret and analyze visual information. This article explores the fundamental principles, techniques, applications, challenges, and future prospects of machine vision. It delves into the key components of machine vision systems, including illumination, optics, sensors, vision processing, and communication systems. The integration of these components forms the backbone of machine vision, enhancing automation, decision-making, and analysis in diverse industries. The article also provides a comprehensive overview of the working mechanism, emphasizing the step-by-step process from image acquisition to decision-making and integration with automation systems. Understanding machine vision's intricate components and working processes is essential for harnessing its capabilities to enhance manufacturing, inspection, and overall industrial operations.

Machine vision, a subset of artificial intelligence and computer vision, has transformed various industries by facilitating the interpretation and analysis of visual information by machines. The capacity to perceive and comprehend the surrounding environment is an essential human characteristic, and machine vision aims to replicate this ability in machines. By utilizing advanced algorithms, machine vision systems can extract valuable insights from images or videos, thereby enabling automation, decision-making, and sophisticated analysis. This article explores the intricacies of machine vision, including its fundamental principles, techniques, applications, challenges, and future prospects. Machine vision systems encompass a collection of integrated components devised to utilize information derived from digital images for the purpose of autonomously directing manufacturing and production activities. These systems also possess the capability to facilitate automated assembly verification and inspection tasks by guiding material handling equipment to position products or materials appropriately within a given process. They find extensive utility across diverse industries and can automate mundane and repetitive responsibilities that would otherwise burden a human inspector or operator. The employment of machine vision systems enables a comprehensive examination of products or parts during the manufacturing process, leading to enhanced yields, decreased defect rates, improved quality, reduced expenses, and heightened consistency in process outcomes.

The key components of machine vision system: Machine vision systems, alternatively known as automated vision systems or vision inspection systems, comprise multiple components that are typically shared among various systems. While each of these components serves a unique purpose on its own and can be found in other systems independently, when operating in tandem, they fulfill specific roles within a machine vision system. (For further details on selecting the hardware components of a machine vision system, please refer to our informative guide titled "Machine Vision System Hardware Selection"). Machine vision systems can be constructed using separate components or can be integrated into a unified unit, such as a smart camera, which consolidates the functionalities of the individual elements into a single package. Additional insights on discrete versus integrated machine vision systems can be accessed in our informative guides titled "PC-based Machine Vision Versus Smart Camera Systems" and "Smart Camera Integration in Machine Vision Systems". Irrespective of whether a discrete or integrated system is utilized, the effectiveness of the machine vision system is also influenced by the characteristics of the parts being assessed. A higher level of consistency in part placement and orientation results in improved system performance.

Typically, machine vision systems consist of five fundamental elements, which encompass:

a. The illumination/ lighting system

b. The optical system or lens

c. The sensor

d. The vision processing system

e. The communication system

Lighting system:
Choosing appropriate lighting for a machine vision system involves maximizing contrast for specific features of interest while minimizing contrast for other features of the part. This may require adjusting the light intensity, selecting the lighting style (such as dome or ring light), and positioning the light source relative to the part and optical system. Modifying these parameters can significantly enhance the system's ability to consistently identify and measure the targeted part features. Lighting options include LED lighting and strobe lights, which enable fast shutter speed image capture.

Optical System:
The optical components of a machine vision system typically consist of a lens or camera integrated with other elements like the sensor. The lens selection determines the field of view, depth of focus, and focal point, which affect the system's ability to observe part features. The optical system may feature interchangeable lenses or fixed lenses in designs that utilize a smart camera. Longer focal length lenses provide higher magnification but reduce the field of view. The selection of the optical system depends on the machine vision system's specific function and the dimensions of the observed feature. The optical system may also have color recognition capability.

Sensors in machine vision systems capture light from the optical system and convert it into digital images. These sensors employ CMOS or CCD technology to produce pixel-based representations of the observed part. The sensor's resolution indicates the number of pixels in the digital image, with higher resolutions offering better image quality and improved ability to discern details. The sensor's resolution is influenced by the size of the observed parts, the dimensions of measurements, tolerances, and other application parameters. Higher resolutions enhance the accuracy of measurements made by the machine vision system.

Vision Processing system:
The vision processing element of a machine vision system takes data from the digital image and utilizes software to perform specific functions that evaluate the observed part. These evaluations involve predefined criteria for accepting or rejecting the part. The steps performed by the vision processing system include acquiring the digital image, preprocessing it as needed for optimal measurements, analyzing the image to locate specific features, collecting measurements, comparing values against dimensional criteria, and determining a pass-fail or go/no-go result for the part.

Communications system:
After the vision processing element completes its steps, the final component of the machine vision system is the communications protocol. This element provides standardized output in a usable format, enabling communication with other components in the production process. Standardized outputs include discrete I/O signals or serial data (such as RS-232 or Ethernet) that can be sent to logging devices or other systems. Discrete I/O signals can control stack lights or actuate solenoid-driven actuators to remove rejected parts from the production pathway. Serial data feeds may be directed to HMI screens for operator information display. System integrators can assist in incorporating communication signals between machine vision systems and other machines in the production cell. Working of machine vision system:To comprehend machine vision, one must understand its foundation in computer vision, which concentrates on the development of algorithms and methods for extracting information from visual data. It encompasses processes like image acquisition, image processing, feature extraction, object recognition, and scene comprehension. Machine vision expands upon these capabilities by incorporating decision-making and analysis, empowering machines to interpret and comprehend visual data within a specific context. Machine vision systems employ hardware components such as cameras and sensors, alongside software algorithms, to acquire, process, and analyse visual information.

1. Image Acquisition: At the onset of machine vision, the first step is image acquisition, which encompasses the capturing of visual data using cameras or sensors. It is imperative to obtain high-quality images to ensure precise analysis and interpretation. To acquire images from diverse angles and lighting conditions, a range of hardware components, including digital cameras, scanners, and depth sensors, are utilized.

2. Image Pre-processing: After image acquisition, it is common practice to subject the acquired images to pre-processing to improve their quality and enable subsequent analysis. This stage involves various tasks, including reducing noise, enhancing the image, and normalizing it. Techniques such as filtering, histogram equalization, and image resizing are employed to eliminate distortions, address lighting inconsistencies, and normalize the images for further processing.

3. Feature Extraction: Feature extraction holds immense importance in machine vision as it entails the identification and extraction of pertinent information from images. Features can range from basic attributes like edges, corners, or textures to more intricate elements such as shapes, objects, or patterns. Feature extraction algorithms, such as edge detection, corner detection, and scale-invariant feature transform (SIFT), are utilized to identify distinctive attributes that facilitate subsequent analysis and recognition.

4. Object Recognition: Object recognition holds significant importance in machine vision as it encompasses the identification and classification of objects within images or video streams. This task enables machines to comprehend the contents of an image and categorize objects into predetermined classes or detect anomalies. Object recognition algorithms, including convolutional neural networks (CNNs), support vector machines (SVMs), and deep learning techniques, play a pivotal role in training models to accurately recognize objects.

5. Scene Understanding: Scene understanding surpasses mere object recognition and entails grasping the context and relationships among objects within a given scene. It necessitates machines to deduce spatial arrangements, object interactions, and semantic significance. Techniques such as semantic segmentation, scene graph analysis, and context modeling empower machines to comprehend intricate scenes, assisting in tasks such as scene interpretation, navigation, and decision-making that is aware of the surrounding context.

6. Decision-Making and Analysis: Machine vision systems amalgamate the ability to make decisions and conduct analyses, enabling machines to make informed assessments based on visual information. By employing algorithms and statistical models, machines can examine patterns, trends, and deviations within images or video streams. This facilitates applications such as quality control, defect detection, object tracking, and predictive maintenance.

7. Integration with Automation Systems: Integration of machine vision with automation systems empowers machines to carry out tasks independently. By incorporating machine vision into robotic systems or industrial machinery, automation can be achieved in processes like sorting, assembly, inspection, and packaging. Machine vision systems offer real-time feedback and control, guaranteeing accuracy and effectiveness in industrial operations.