Making Machine See Notes – Class 12 AI (843) | Unit-3 Computer Vision

Master every concept from Unit-3 Making Machines See with these Class 12 AI notes. We’ve simplified Computer Vision as per the latest CBSE curriculum into clear, easy-to-understand points designed specifically for your board exams. Simple explanations, focused content—everything you need to score full marks.

Computer Vision is a branch of Artificial Intelligence (AI) that uses sensing devices and deep learning techniques to enable machines to understand and interpret visual information.

What is Computer Vision?

• It is similar to human vision, using cameras, data, and algorithms like the human retina, optic nerves, and visual cortex.
• CV extracts meaningful information from images, videos, and other visual inputs.
• Computer Vision systems are used to inspect products, monitor infrastructure, and analyze production processes in real-time.
• Deep learning models in CV achieve high accuracy in tasks like facial recognition, object detection, and image classification.

Working Of Computer Vision

• Computer Vision focuses on processing and analysing digital images and videos to understand their content.
• Understanding digital images is a fundamental part of computer vision.

Basics of Digital Images

• A digital image is a picture stored as a sequence of numbers that computer can interpret.
• Digital images can be created by several ways such as
  • Design software (e.g., Paint, Photoshop)
  • Digital cameras
  • Scanner

Interpretation of Image in Digital Form

• A computer sees an image as a collection of tiny units called pixels (picture elements).
• Each pixel represents a specific color value.  These pixels together form the digital image.
• During digitization, an image is converted into a grid of pixels.
• Resolution of image is determined by number of pixels:
  • More pixels → Higher resolution → More detail
• In monochrome (black & white) images: Pixel values range from 0 to 255 where 0 = Black and 255 = White

Computer Vision – Process:

Computer Vision process involves five stages which are explained below.

Image Acquisition

• Image Acquisition is the first stage.
• It involves capturing digital images or videos and provides the raw data for further analysis.
• The quality of acquired images affects further processing and analysis.
• Imaging device capabilities and resolution impact image quality.
• Higher resolution captures more details and clearer images.
• Image acquisition is affected by factors like:
  • Lighting conditions
  • Angle of capture
• In scientific and medical fields, advanced techniques are used. For example: MRI and CT scans

Preprocessing:

• Preprocessing in Computer Vision improves the quality of the acquired image.
• It uses various techniques to make images suitable for analysis.

Common Preprocessing Techniques

• Noise Reduction
  • Removes unwanted elements like blurriness, spots, or distortions and makes the image clearer.
  • Example: Removing grainy effects in low-light images.
• Image Normalization
  • Standardizes pixel values across images. Adjusts values to a fixed range (e.g., 0–1 or -1 to 1).
  • Ensures consistency in datasets.
  • Example: Converting pixel values from 0–255 to 0–1.
• Resizing/Cropping
  • Changes image size or aspect ratio and ensures all images have the same dimensions.
  • Example: Resizing images to 224 × 224 pixels.
• Histogram Equalization
  • Adjusts brightness and contrast.
  • Spreads pixel intensity evenly and enhances details in dark or bright areas.
  • Example: Improving low-contrast images.
• Goal of Preprocessing
  • Remove noise or disturbances.
  • Highlight important features.
  • Maintain consistency and uniformity in images.

Feature Extraction

• Feature Extraction involves identifying and extracting important visual patterns or attributes from a pre-processed image.
• The methods used for feature extraction depend on:
  • Complexity of the image
  • Available computational resources
  • Requirements of the application
• Types of Feature Extraction
  • Edge Detection
    • Identifies boundaries between regions.
    • Detects sharp changes in intensity.
  • Corner Detection
    • Identifies points where two or more edges meet.
    • Focuses on areas with high curvature or sharp changes.
  • Texture Analysis
    • Extracts patterns like smoothness, roughness, or repetition.
  • Colour-Based Feature Extraction
    • Analyses colour distribution in an image.
    • Helps differentiate objects based on colour.
• In deep learning approaches, feature extraction is done automatically using Convolutional Neural Networks (CNNs) during training.

Detection/Segmentation

• Detection/Segmentation focuses on identifying objects or regions of interest in an image.
• It is important for applications like:
  • Autonomous driving
  • Medical imaging
  • Object tracking
• This stage is divided into:
  • Single Object Tasks
  • Multiple Object Tasks

Single Object Tasks

• Focus on analysing or identifying a single object in an image with two main objectives:
• Classification
  • Determines the category or class of an object and help identify the object. It can uses algorithms KNN (K-Nearest Neighbour) – supervised or K-means clustering – unsupervised
• Classification + Localization
  • Classifies the object and also locates it. It uses bounding boxes to show object position.

Multiple Object Tasks

• Multiple Object Tasks deal with images containing multiple objects or different classes.
• Aim to identify and distinguish between various objects in an image.

Types of Multiple Object Tasks

• Object Detection
  • Identifies and locates multiple objects in an image.
  • Draws bounding boxes around each object.Assigns class labels to detected objects.Uses features and learning algorithms for recognition.
  • Common algorithms: R-CNN, R-FCN, YOLO (You Only Look Once), SSD (Single Shot Detector)
• Image Segmentation
  • Image Segmentation creates masks around pixels with similar characteristics.
  • It assigns a class to pixels in the input image.
  • Helps in understanding the image at a detailed (granular) level.
  • Each object is represented using a pixel-wise mask.
  • Makes it easier to identify and separate different objects.Uses techniques like edge detection (detecting changes in brightness).
  • There are different types of image segmentation available.
• Types of Image Segmentation
  • Semantic Segmentation
    • Classifies pixels of the same class together.
    • Does not differentiate between objects of the same class.
  • Instance Segmentation
    • Classifies pixels for each individual object.
    • Differentiates between objects even if they belong to the same class.
• High-Level Processing
  • High-Level Processing is the final stage of computer vision.
  • It interprets and extracts meaningful information from detected objects or regions in images.
  • Helps computers achieve a deeper understanding of visual content.
  • Enables informed decision-making based on visual data.
  • Tasks include:
    • Object recognition
    • Scene understanding
    • Context analysis

Applications of Computer Vision

• Facial Recognition
  • Used by social media platforms to detect and tag users.
• Healthcare
  • Detects diseases and abnormalities.
  • Helps in identifying cancerous tumours.
• Self-Driving Vehicles
  • Understand surroundings using cameras.
  • Detect vehicles, objects, traffic signals, and pedestrians.
• Optical Character Recognition (OCR)
  • Extracts text from images or documents.
  • Used in invoices, bills, and articles.
• Machine Inspection
  • Detects defects and flaws in machines or products.
  • Helps in quality control during manufacturing.
• 3D Model Building
  • Creates 3D models from real-world objects.
  • Used in robotics, autonomous driving, 3D tracking, and AR/VR.
• Surveillance
  • Uses CCTV footage to monitor activities.
  • Identifies suspicious behaviour and prevents crimes.
• Fingerprint Recognition and Biometrics
  • Identifies individuals using fingerprints or biometric data.
  • Used for authentication and security.

Challenges of Computer Vision

• Reasoning and Analytical Issues
  • Requires accurate interpretation, not just image identification.
  • Needs strong reasoning and analytical capabilities.
  • Without these, extracting meaningful insights becomes difficult.
• Difficulty in Image Acquisition
  • Affected by lighting variations, different perspectives, and scales.
  • Complex scenes with multiple objects increase difficulty.Occlusion (objects blocking each other) adds complexity.
  • High-quality image data is necessary for accurate analysis.
• Privacy and Security Concerns
  • Surveillance systems may violate privacy rights.
  • Facial recognition raises ethical issues.
  • Requires careful handling and regulations.
• Duplicate and False Content
  • Risk of fake or misleading images/videos.
  • Exploitation of system vulnerabilities by malicious users.Data breaches can spread duplicate or false content.
  • Leads to misinformation and reputational damage.

Future of Computer Vision

• Computer Vision has evolved from basic image processing to advanced systems with human-like precision.
• Growth is driven by:
  • Advances in deep learning algorithms
  • Availability of large amounts of labelled data
• Machines can now perceive and analyse images and videos effectively.
• Future possibilities of computer vision are vast and impactful.
• It will have a significant and far-reaching impact on societ
• Computer vision can bring transformative benefits for humanity.