KEYWORDS: Cameras, Imaging systems, Machine vision, Light sources and illumination, Digital cameras, Image resolution, Spatial resolution, Signal to noise ratio
This paper preliminarily investigates the performance of machine vision system from the perspective of contrast and object detection process. We set up a testing system using a lightbox, transmissive/reflective test charts, a photometer, and two cameras. The photometer was used to obtain the standard luminance of the test chart. First, we obtain the DN response characteristics for luminance of two cameras with different dynamic ranges (72 dB and 123.6 dB). Based on this result the relationship between luminance domain and camera domain contrast ratio is provided. Distribution of the signal and contrast in the camera domain under low luminance conditions show the advantage of 16bit camera over 8bit camera. To link the camera's performance in practical scene, we conducted imaging tests of reflective resolution targets under various illuminance levels. We observed that the contrast and imaging quality of resolution targets by the camera at critical states can help establish correlation between single-metrics and scene-based imaging recognition performance evaluation.
Machine vision is not a mere upgrade of the specification of the current imaging devices, but rather a form of visual perception technology that involves intelligent modules in the processes of measurement, processing, and decision- making. Given the novel functionalities and features of machine vision-based intelligent detection devices, the traditional evaluation methods based on testing the physical parameters of imaging devices need further refinement and development. Taking the electroluminescence (EL) imaging in photovoltaic (PV) tests as an example, we investigate the influence of changes in dataset characteristics on the performance of object detection by combining digital image processing and deep learning methods. Features regarding to the crack-type defect datasets, such as the grayscale, contrast, shape and resolution, are controlled and adjusted based on new generated datasets from the original datasets. From the numerical experiments, some new aspects for evaluating the intelligent detection.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.