SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
A compact VLSI neural computer integrated with an active pixel sensor has been under development to mimic what is inherent in biological visio systems.
We present an algoritm for detecting a specified set of targets for an Automatic Target Recognition (ATR) application.
We overview our recent progress in lander-based robotics for Mars planetary science. Utilizing a 1:1 scale laboratory replica of the NASA Mars Surveyor '98 mission, JPL engineers and Mars sciencce colleagues have demonstrated approaches to lander science functions such as robotic sample acquisiiton and deposition, end-effector based microscopic viewing, hand-carried science instrument data collection, and science instrument emplacement by a robot.
Explore the source record for details and available documents.
Amherst Systems manufactures foveal machine vision technology and systems commercially available to end-users and system integrators. This technology was initially developed under NASA contracts NAS9-19335 (Johnson Space Center) and NAS1-20841 (Langley Research Center). This technology is currently being delivered to university research facilities and military sites. More information may be found in www.amherst.com.
Automated Camera Calibration (ACAL) is a computer program that automates the generation of calibration data for camera models used in machine vision systems. Machine vision camera models describe the mapping between points in three-dimensional (3D) space in front of the camera and the corresponding points in two-dimensional (2D) space in the camera s image. Calibrating a camera model requires a set of calibration data containing known 3D-to-2D point correspondences for the given camera system. Generating calibration data typically involves taking images of a calibration target where the 3D locations of the target s fiducial marks are known, and then measuring the 2D locations of the fiducial marks in the images. ACAL automates the analysis of calibration target images and greatly speeds the overall calibration process.
Machine vision research is examined as it relates to the NASA Space Station program and its associated Orbital Maneuvering Vehicle (OMV). Initial operation of OMV for orbital assembly, docking, and servicing are manually controlled from the ground by means of an on board TV camera. These orbital operations may be accomplished autonomously by machine vision techniques which use the TV camera as a sensing device. Classical machine vision techniques are described. An alternate method is developed and described which employs a syntactic pattern recognition scheme. It has the potential for substantial reduction of computing and data storage requirements in comparison to the Two-Dimensional Fast Fourier Transform (2D FFT) image analysis. The method embodies powerful heuristic pattern recognition capability by identifying image shapes such as elongation, symmetry, number of appendages, and the relative length of appendages.
The figure presents selected views of a compact microscope imaging system (CMIS) that includes a miniature video microscope, a Cartesian robot (a computer- controlled three-dimensional translation stage), and machine-vision and control subsystems. The CMIS was built from commercial off-the-shelf instrumentation, computer hardware and software, and custom machine-vision software. The machine-vision and control subsystems include adaptive neural networks that afford a measure of artificial intelligence. The CMIS can perform several automated tasks with accuracy and repeatability . tasks that, heretofore, have required the full attention of human technicians using relatively bulky conventional microscopes. In addition, the automation and control capabilities of the system inherently include a capability for remote control. Unlike human technicians, the CMIS is not at risk of becoming fatigued or distracted: theoretically, it can perform continuously at the level of the best human technicians. In its capabilities for remote control and for relieving human technicians of tedious routine tasks, the CMIS is expected to be especially useful in biomedical research, materials science, inspection of parts on industrial production lines, and space science. The CMIS can automatically focus on and scan a microscope sample, find areas of interest, record the resulting images, and analyze images from multiple samples simultaneously. Automatic focusing is an iterative process: The translation stage is used to move the microscope along its optical axis in a succession of coarse, medium, and fine steps. A fast Fourier transform (FFT) of the image is computed at each step, and the FFT is analyzed for its spatial-frequency content. The microscope position that results in the greatest dispersal of FFT content toward high spatial frequencies (indicating that the image shows the greatest amount of detail) is deemed to be the focal position.
Battery cell defect classification is primarily done manually by a human conducting a visual inspection to determine if the battery cell is acceptable for a particular use or device. Human visual inspection is a time consuming task when compared to an inspection process conducted by a machine vision system. Human inspection is also subject to human error and fatigue over time. We present a machine vision technique that can be used to automatically identify defective sections of battery cells via a morphological feature-based classifier using an adaptive two-dimensional fast Fourier transformation technique. The initial area of interest is automatically classified as either an anode or cathode cell view as well as classified as an acceptable or a defective battery cell. Each battery cell is labeled and cataloged for comparison and analysis. The result is the implementation of an automated machine vision technique that provides a highly repeatable and reproducible method of identifying and quantifying defects in battery cells.
A unique and customizable machinevision and image-data-processing technique has been developed for use in automated identification of cells that are optimal for patch clamping. [Patch clamping (in which patch electrodes are pressed against cell membranes) is an electrophysiological technique widely applied for the study of ion channels, and of membrane proteins that regulate the flow of ions across the membranes. Patch clamping is used in many biological research fields such as neurobiology, pharmacology, and molecular biology.] While there exist several hardware techniques for automated patch clamping of cells, very few of those techniques incorporate machine vision for locating cells that are ideal subjects for patch clamping. In contrast, the present technique is embodied in a machine-vision algorithm that, in practical application, enables the user to identify good and bad cells for patch clamping in an image captured by a charge-coupled-device (CCD) camera attached to a microscope, within a processing time of one second. Hence, the present technique can save time, thereby increasing efficiency and reducing cost. The present technique involves the utilization of cell-feature metrics to accurately make decisions on the degree to which individual cells are "good" or "bad" candidates for patch clamping. These metrics include position coordinates (x,y) in the image plane, major-axis length, minor-axis length, area, elongation, roundness, smoothness, angle of orientation, and degree of inclusion in the field of view. The present technique does not require any special hardware beyond commercially available, off-the-shelf patch-clamping hardware: A standard patchclamping microscope system with an attached CCD camera, a personal computer with an imagedata- processing board, and some experience in utilizing imagedata- processing software are all that are needed. A cell image is first captured by the microscope CCD camera and image-data-processing board, then the image data are analyzed by software that implements the present machine-vision technique. This analysis results in the identification of cells that are "good" candidates for patch clamping (see figure). Once a "good" cell is identified, a patch clamp can be effected by an automated patchclamping apparatus or by a human operator. This technique has been shown to enable reliable identification of "good" and "bad" candidate cells for patch clamping. The ultimate goal in further development of this technique is to combine artificial-intelligence processing with instrumentation and controls in order to produce a complete "turnkey" automated patch-clamping system capable of accurately and reliably patch clamping cells with a minimum intervention by a human operator. Moreover, this technique can be adapted to virtually any cellular-analysis procedure that includes repetitive operation of microscope hardware by a human.
A mathematical model of image artifacts produced by dust particles on lenses has been derived. Machine-vision systems often have to work with camera lenses that become dusty during use. Dust particles on the front surface of a lens produce image artifacts that can potentially affect the performance of a machine-vision algorithm. The present model satisfies a need for a means of synthesizing dust image artifacts for testing machine-vision algorithms for robustness (or the lack thereof) in the presence of dust on lenses. A dust particle can absorb light or scatter light out of some pixels, thereby giving rise to a dark dust artifact. It can also scatter light into other pixels, thereby giving rise to a bright dust artifact. For the sake of simplicity, this model deals only with dark dust artifacts. The model effectively represents dark dust artifacts as an attenuation image consisting of an array of diffuse darkened spots centered at image locations corresponding to the locations of dust particles. The dust artifacts are computationally incorporated into a given test image by simply multiplying the brightness value of each pixel by a transmission factor that incorporates the factor of attenuation, by dust particles, of the light incident on that pixel. With respect to computation of the attenuation and transmission factors, the model is based on a first-order geometric (ray)-optics treatment of the shadows cast by dust particles on the image detector. In this model, the light collected by a pixel is deemed to be confined to a pair of cones defined by the location of the pixel s image in object space, the entrance pupil of the lens, and the location of the pixel in the image plane (see Figure 1). For simplicity, it is assumed that the size of a dust particle is somewhat less than the diameter, at the front surface of the lens, of any collection cone containing all or part of that dust particle. Under this assumption, the shape of any individual dust particle artifact is the shape (typically, circular) of the aperture, and the contribution of the particle to the attenuation factor for a given pixel is the fraction of the cross-sectional area of the collection cone occupied by the particle. Assuming that dust particles do not overlap, the net transmission factor for a given pixel is calculated as one minus the sum of attenuation factors contributed by all dust particles affecting that pixel. In a test, the model was used to synthesize attenuation images for random distributions of dust particles on the front surface of a lens at various relative aperture (F-number) settings. As shown in Figure 2, the attenuation images resembled dust artifacts in real test images recorded while the lens was aimed at a white target.
An astronaut's helmet is an invariant, rigid image element that is well suited for identification and tracking using current machine vision technology. Future space exploration will benefit from the development of astronaut detection software for search and rescue missions based on EVA helmet identification. However, helmets are solid white, except for metal brackets to attach accessories such as supplementary lights. We compared the performance of a widely used machine vision pipeline on a standard-issue NASA helmet with and without affixed experimental feature-rich patterns. Performance on the patterned helmet was far more robust. We found that four different feature-rich patterns are sufficient to identify a helmet and determine orientation as it is rotated about the yaw, pitch, and roll axes. During helmet rotation the field of view changes to frames containing parts of two or more feature-rich patterns. We took reference images in these locations to fill in detection gaps. These multiple feature-rich patterns references added substantial benefit to detection, however, they generated the majority of the anomalous cases. In these few instances, our algorithm keys in on one feature-rich pattern of the multiple feature-rich pattern reference and makes an incorrect prediction of the location of the other feature-rich patterns. We describe and make recommendations on ways to mitigate anomalous cases in which detection of one or more feature-rich patterns fails. While the number of cases is only a small percentage of the tested helmet orientations, they illustrate important design considerations for future spacesuits. In addition to our four successful feature-rich patterns, we present unsuccessful patterns and discuss the cause of their poor performance from a machine vision perspective. Future helmets designed with these considerations will enable automated astronaut detection and thereby enhance mission operations and extraterrestrial search and rescue.
Early fire detection in industrial environments is critical to preventing equipment damage, personal injury, and operational disruptions. Traditional smoke detectors, while effective, often experience delays due to the time required for smoke to reach sensors, allowing fires to spread. Manual fire watch operations and human surveillance of camera feeds are resource-intensive and prone to human error. To address these challenges, this paper explores the application of convolutional neural networks for automated fire detection, specifically in industrial settings. By leveraging 11 different pre-trained machine vision models from TensorFlow and enhancing them with transfer learning on a custom-built industrial fire dataset, we optimized fire detection performance. Here, we analyzed each machine vision model architecture in terms of its depth, width, and input image resolution, considering both resource requirements and detection accuracy. We further explored the option of combining multiple models into an ensemble classifier to evaluate whether the performance improvements could justify the much greater computational complexity and other practical impacts. A cost-benefit analysis is presented to evaluate the trade-offs between performance and computational expense. Our findings identify that EfficientNetV2L, specifically tailored for industrial applications, provides the optimal balance between costs involved in training and using the model versus the overall fire detection performance. Additionally, we present a qualitative analysis of model performance using the technique of gradient-based class activation mapping to provide explainability by visualizing model decisions.
Spatial displays, and a constraint that they do not place on the use of spatial instruments are discussed. Much of the work done in visual perception by psychologists and by computer scientists has concerned displays that show the motion of rigid objects. Typically, if one assumes that objects are rigid, one can then proceed to understand how the constant shape of the object can be perceived (or computed) as it moves through space. The author maintains that photographs and cinema are visual displays that are also powerful forms of art. Their efficacy, in part, stems from the fact that, although viewpoint is constrained when composing them, it is not nearly so constrained when viewing them. It is obvious, according to the author, that human visual systems did not evolve to watch movies or look at photographs. Thus, what photographs and movies present must be allowed in the rule-governed system under which vision evolved. Machine-vision algorithms, to be applicable to human vision, should show the same types of tolerance.
This paper presents on-going research in robotic inspection of space platforms. Three main areas of investigation are discussed: machine vision inspection techniques, an integrated sensor end-effector, and an orbital environment laboratory simulation. Machine vision inspection utilizes automatic comparison of new and reference images to detect on-orbit induced damage such as micrometeorite impacts. The cameras and lighting used for this inspection are housed in a multisensor end-effector, which also contains a suite of sensors for detection of temperature, gas leaks, proximity, and forces. To fully test all of these sensors, a realistic space platform mock-up has been created, complete with visual, temperature, and gas anomalies. Further, changing orbital lighting conditions are effectively mimicked by a robotic solar simulator. In the paper, each of these technology components will be discussed, and experimental results are provided.
The development of generalized contour/texture discrimination techniques is a central element necessary for machine vision recognition and interpretation of arbitrary images. Here, the visual perception of texture, selected studies of texture analysis in machine vision, and diverse small samples of contour and texture are all used to provide insights into the fundamental characteristics of contour and texture. From these, an experimental discrimination scheme is developed and tested on a battery of natural images. The visual perception of texture defined fine texture as a subclass which is interpreted as shading and is distinct from coarse figural similarity textures. Also, perception defined the smallest scale for contour/texture discrimination as eight to nine visual acuity units. Three contour/texture discrimination parameters were found to be moderately successful for this scale discrimination: (1) lightness change in a blurred version of the image, (2) change in lightness change in the original image, and (3) percent change in edge counts relative to local maximum.
Recent proliferation of small Unmanned Aerial Systems (sUAS) applications requires onboard collision avoidance systems to mitigate the risk of collision with non-cooperative aircraft and manned aircraft, which may not see sUAS in time to perform an avoidance maneuver. An attractive avenue for onboard collision avoidance is the utilization of machine vision cameras due to their low size, weight and power (SWaP) requirements. In this paper, we characterize the range performance of a machine vision system developed in-house and mounted onto an sUAS. The technique was designed to estimate the performance of a sense-and-avoid system to ensure that the sensing components meet the well-clear requirements for the chosen platform and avoidance strategy. Experimental flight-test data was acquired from test-flights flown along multiple collision geometries for two intruders: a general Aviation (GA) aircraft and a fixed-wing sUAS. The ownship and both intruders were instrumented with inertial navigation systems (INS) recording position and attitude information. The range at first detection, 𝑹𝟎, was extracted from in-flight imagery of head-on collision course geometry synchronized with INS data from both aircraft and ground-truth values extracted from the raw imagery. This initial detection distance, 𝑹𝟎, scales with atmospheric attenuation. Therefore, under clear sky conditions, the derived 𝑹𝟎 value represents the upper bound on the detection range achievable by the test configuration of the detector. Results indicate that the maximum initial detection distance for a 4k resolution action camera fitted with a 41º Field of View (FOV) lens is 2.763 ± 0.037 km for a GA aircraft and 0.881 ± 0.061 km for a fixed-wing sUAS, respectively. The in results this study suggest that a vision-based detect and track system may be analyzed using the sensor characterization and contextualized within aircraft well-clear volumes.