Zebra Aurora Imaging Library Software

• Solve applications rather than develop underlying tools by leveraging a toolkit with a more than 25-year history of reliable performance.
• Tackle applications with utmost confidence using field-proven tools for analyzing, classifying, locating, measuring, reading, and verifying
• Base analysis on monochrome and color 2D images as well as 3D profiles, depth maps, and point clouds
• Harness the full power of today’s hardware through optimizations exploiting SIMD, multi-core CPU, and multi-CPU technologies
• Support platforms ranging from smart cameras to high-performance computing (HPC) clusters via a single consistent and intuitive application programming interface (API)
• Obtain live data in different ways, with support for analog, Camera Link®, CoaXPress®, DisplayPort™, GenTL, GigE Vision®, HDMI™, SDI, Linux® and USB3 Vision® interfaces
• Maintain flexibility and choice by way of support for 64-bit Windows® and Linux® along with Intel® and Arm® processor architectures
• Leverage available programming know-how with support for C, C++, C# and CPython languages
• Experiment, prototype, and generate program code using Aurora Imaging Library CoPilot interactive environment
• Increase productivity and reduce development costs with Vision Academy online and on-premises training

Industrial imaging tools Aurora Imaging Library1, formerly Matrox Imaging Library, is a comprehensive collection of software tools for developing machine vision and image analysis applications. Aurora Imaging Library includes tools for every step in the process, from application feasibility to prototyping, through to development and ultimately deployment. The software development kit (SDK) features interactive software and programming functions for image capture, and archiving. These tools are designed to enhance productivity, thereby reducing the time and effort required to bring solutions to market. Image capture, and analysis operations have the accuracy and robustness needed to tackle the most demanding applications. These operations are also carefully optimized for speed to address the severe time constraints encountered in many applications. Aurora Imaging Library development First released in 1993, Aurora Imaging Library has evolved to keep pace with and foresee emerging industry requirements. It was conceived with an easy-to-use, coherent API that has stood the test of time. Aurora Imaging Library pioneered the concept of hardware independence with the same API for different image acquisition and processing platforms. A team of dedicated, highly skilled computer scientists, mathematicians, software engineers, and physicists continue to maintain and enhance Aurora Imaging Library. Aurora Imaging Library is maintained and developed using industry recognized best practices, including peer review, user involvement, and daily builds.

Overview

Aurora Imaging Library is a comprehensive collection of software tools for developing machine vision and image analysis applications. It includes interactive software and programming functions for image capture, and archiving. These tools are designed to enhance productivity and reduce the time and effort required to bring solutions to market. Aurora Imaging Library is maintained and developed using industry-recognized best practices, and daily builds.

Deep learning model for object detection

Aurora Imaging Library includes Classification tools for automatically categorizing image content or previously extracted features using machine learning. Image-oriented classification makes use of deep learning technology—specifically the convolutional neural network (CNN) and variants—in three distinct approaches. The global approach assigns images or image regions to preestablished classes. It lends itself to identification tasks where the goal is to distinguish between similarly looking objects including those with slight imperfections. The results for each image or image region consist of the most likely class and a score for each class. The segmentation approach generates maps indicating the preestablished class and score for all image pixels. It is appropriate for detection tasks where the objective is to determine the incidence, location, and extent of flaws or features. These features can then be further analyzed and measured using traditional tools like Blob Analysis. The object detection approach locates instances of preestablished classes. It is suited for inspection tasks whose goal is to find, size and count objects or features. The result for each located instance is the most likely class, the score and a bounding box including the corner coordinates, center, height and width. Image-oriented classification is particularly well-suited for inspecting images of highly textured, naturally varying, and acceptably deformed goods in complex and varying scenes. Users can opt to train a deep neural network on their own or commission Zebra to do so using previously collected images; these images must be both adequate in number and representative of the expected application conditions. Different types of training are supported, such as transfer learning and fine-tuning, all starting from one of the supplied pre-defined deep neural network architectures.

Deep learning inference using an Intel integrated GPU and NVIDIA GPU

Aurora Imaging Library includes deep learning inference using an Intel integrated GPU and NVIDIA GPU.

3D surface matcher

Aurora Imaging Library includes a tool for finding a surface model—including multiple occurrences—at wide-ranging orientations in a point cloud. A surface model is defined from a point cloud obtained from a 3D camera or sensor, or from a CAD (PLY or STL) file. Various controls are provided to influence the search accuracy, robustness, and speed. Search results include the number of occurrences found and for each occurrence, the score, error, number of points, center coordinates, and estimated pose. The 3D surface matcher is suited for locating and inspecting objects in a point cloud. It makes it possible to segment a point cloud into blobs, calculate numerous blob features, filter and sort blobs by features, as well as select and combine blobs. A point cloud can subsequently be analyzed using Aurora Imaging Library’s 2D vision tools like Pattern Recognition—without being affected by illumination variations or surface texture—and Character Recognition, when the alphanumeric code to read protrudes from, but has the same color as the background. A profile or cross section can be analyzed using Metrology. The 3D surface matcher is particularly well suited for inspecting images of highly textured, and acceptably deformed goods in complex and varying scenes.

3D shape finding tool

Aurora Imaging Library provides a tool for locating specific shapes—cylinders, (hemi)spheres, rectangular planes, and boxes—in a point cloud. The shape to find is specified either numerically or from one or two previously defined elementary objects. Several settings are provided to tune the finding process accuracy, and speed. Results include the number of occurrences found and for each occurrence, and center coordinates. Additional results include the radius for spheres and cylinders, length(s) for cylinders, and boxes, central axis and bases for cylinders, normal unit vector for rectangular planes, and number of visible faces for boxes.

3D blob analysis tool

Aurora Imaging Library provides a 3D Blob Analysis tool for locating and inspecting objects in a point cloud. It makes it possible to segment a point cloud into blobs, as well as select and combine blobs. The 3D Blob Analysis tool is particularly well suited for inspecting images of highly textured, and acceptably deformed goods in complex and varying scenes.

Feature extraction and analysis tools

Aurora Imaging Library provides a choice of tools for image analysis: Blob Analysis and Edge Finder. These tools are used to identify and measure basic features for determining object presence and location, and to further examine objects. The Blob Analysis tool works on segmented binary images, where objects are previously separated from the background and one another. The tool—using run-length encoding—quickly identifies blobs and can measure over 50 binary and grayscale characteristics. Measurements can be used to sort and select blobs. The tool also reconstructs and merges blobs, which is useful when working with blobs that straddle successive images. The Edge Finder tool is well suited for scenes with changing, uneven illumination. The tool—using gradient-based and Hessianbased approaches—quickly identifies contours, as well as crests or ridges, in monochrome or color images and can measure over 50 characteristics with sub-pixel accuracy. Measurements can be used to sort and select edges. The edge extraction method can be adjusted to tailor performance.

Image-oriented classification (global approach)

Aurora Imaging Library includes Classification tools for automatically categorizing image content or previously extracted features using machine learning. Image-oriented classification makes use of deep learning technology—specifically the convolutional neural network (CNN) and variants—in three distinct approaches. The global approach assigns images or image regions to preestablished classes. It lends itself to identification tasks where the goal is to distinguish between similarly looking objects including those with slight imperfections. The results for each image or image region consist of the most likely class and a score for each class. The segmentation approach generates maps indicating the preestablished class and score for all image pixels. It is appropriate for detection tasks where the objective is to determine the incidence, and acceptably deformed goods in complex and varying scenes.

Image-oriented classification (segmentation approach)

Aurora Imaging Library includes Classification tools for automatically categorizing image content or previously extracted features using machine learning. Image-oriented classification makes use of deep learning technology—specifically the convolutional neural network (CNN) and variants—in three distinct approaches. The