Uncalibrated stereo imagery experimental and analytical results are presented for path planning and navigation. An
Army Research and Development Engineering Command micro-size UAV was outfitted with two commercial cameras
and flown over varied landscapes. Polaris Sensor Technologies processed the data post flight with an image
correspondence algorithm of their own design. Stereo disparity (depth) was computed despite a quick assembly, image
blur, intensity saturation, noise and barrel distortion. No camera calibration occurred. Disparity maps were computed at
a processing rate of approximately 5 seconds per frame to improve perception. Disparity edges (treeline to ground, voids
and plateaus) were successfully observed and confirmed to be properly identified. Despite the success of localizing this
disparity edges sensitivity to saturated pixels, lens distortion and defocus were strong enough to overwhelm more subtle
features such as the contours of the trees, which should be possible to extract using this algorithm. These factors are
being addressed. The stereo data is displayed on a flat panel 3D display well suited for a human machine interface in
field applications. Future work will entail extraction of intelligence from acquired data and the overlay of such data on
the 3D image as displayed.
Polaris Sensor Technologies, Inc. is identifying target pixels in IR imagery at signal to noise (SNR) ranges from 1.25 to
3 with a mixed set of algorithms that are candidates for next generation focal planes. Some of these yield less than 50
false targets and a 95% probability of detection in this low SNR range. What has been discovered is that single frame
imagery combined with IMU data can be input into a host of algorithms like Neural Networks and filters to isolate
signals and cull noise. Solutions for nonlinear thresholding approaches can be solved using both genetic algorithms and
neural networks. What is being addressed is how to implement these approaches and apply them to point target
detection scenarios. The large format focal planes will flood the down stream image processing pipelines used in real
time systems, and this team wonders if data can be thinned near the FPA using one of these techniques. Delivering all
the target pixels with a minimum of false positives is the goal addressed by the group. Algorithms that can be digitally
implemented in a ROIC are discussed as are the performance statistics Probability of Detection and False Alarm Rate.
Results from multiple focal planes for varied scenarios will be presented.
Pulse Coupled Neural Networks (PCNNs) have been shown to be of value in image processing applications, especially
at identifying features of small spatial extent at low signal to noise ratio. In our use of the PCNN, every pixel in a scene
feeds a neuron in a fully connected lateral neural network. Nearest neighbor neurons contribute to the output of any
given neuron using weights that link the neuron and its neighborhood in both a linear and a non-linear fashion. The
network is pulsed, and the output of the network at each pulse is a binary mask of neurons that are active. Pulsing drives
the network to evaluate its state. The multi-dimensionality and the non-linear nature of the network make selecting
weights using trial and error a non-trivial problem. It is important that the desired features of the input are identified on
a predictable pulse, a problem that has yet to be sufficiently addressed by proponents of the PCNN. Our method to
overcome these problems is to use a Genetic Algorithm to select the set of PCNN coefficients which will identify the
pixels of interest on a predetermined pulse. This method enables PCNNs to be trained, which is a novel capability and
renders the method of use for applications.
A Pulse Coupled Neural Network (PCNN) has been developed in order to segment image data to reduce the amount of downstream processing. This paper discusses the results of applying the PCNN algorithm to data generated by various sensor platforms. The PCNN algorithm was applied to data generated by a Long Wave Infrared Imaging Polarimeter. The PCNN correctly identified the concealed vehicles and the disturbed earth and rejected 96% of the remaining pixels because they had no information content. Next, the results of applying the PCNN algorithm to noisy infrared seeker data are presented. The PCNN correctly idnetified the target even though the background was quite noisy. Finally, the PCNN algorith was applied to images containing solar glint. It correctly passed only 3& of the pixels to the downstream target/glint decision algorithm. To obtain maximum data throughput, the PCNN can be implemented in hardware.
Adaptive optics systems are used to maintain an optical system at its optimum performance through real time corrections of a wavefront. Deformable mirrors have traditionally been relatively large, expensive devices, suitable for systems such as large telescopes. The objective of the present work is to expand the range of systems that can employ adaptive optics by developing a small, low-cost MEMS deformable mirror. This deformable mirror uses a continuous membrane and has 61 actuators arranged in to approximate a circular pattern. Each actuator has an associated spring suspension, allowing it to push as well as pull on the membrane, producing locally convex or concave curvature. The folded springs are positioned so as to maximize the lateral stability. Maximum actuator displacement is six microns at less than 200 volts. The actuator resonant frequency, is greater than 10 kHz, allowing high-frequency updates of the mirror shape. To operate at high speed, the device must be sealed in a low-pressure environment. Each microactuator uses a vertical comb drive to achieve large travel at a reasonable voltage. The continuous membranes are made of silicon or silicon nitride. Both the actuator and membrane are fabricated with bulk micromachine process technologies. The design targets laser based communication specifications and medical imaging applications.
Pulse couple neural networks (PCNN) have demonstrated some very desirable properties. Chief among these is its ability to segment images very rapidly and very well. This capability has been demonstrated with many different types of imagery including synthetic aperture radar imagery, infrared imagery, optical correlator output imagery, and medical diagnostic imagery. Most of the implementations of this network have been done in software. Several attempts have been made to build electronic versions with varying degrees of success. Recently, an Army Phase II SBIR was awarded to incorporate a PCNN in a smart detector for both military and medical applications. One of the inherent difficulties in building an electronic PCNN is implementing the linking field that is the strength of this network. An optical implementation of the linking would potentially simplify the problem and take advantage of the inherent parallelism of optics. The resultant hardware could be simpler and faster than previous implementations making it an attractive solution. This paper will discuss the current status of the SBIR program, and present possible optical implementations using recently developed Vertical Cavity Surface Emitting Laser arrays.
Digital image interpretation is the basis of medical diagnoses. Through extensive review of patient data, an algorithm was developed to identify features of diagnostic importance in radiological images. The algorithm is generally applicable. Results from cardiac, lung, and military imagery are reported. The algorithm uses a pulse coupled neural network. It is this neural network that is fabricated on a custom CMOS chip. Each neuron of the pulse coupled neural network accepts an external optical input. The optical input is accomplished by a photo-detector. The neurons communicate laterally through a voltage grid. The communication strength, light sensitivity and other global parameters are under external control. A programmable logic array is on the camera board. Data for a specific neuron is accessed by an addressing scheme typically used for a CCD array. The individual neuron speed ranges from 10 to 50 Mhertz, and is fixed by a digital clock. The current chip is configured to operate at 300 Hertz. The chip logic is a hybrid of analog and digital circuitry to minimize the neuron size, maximize the number of neurons at a fixed cost. The hybrid circuitry also minimized the noise level in the chip.
Diffraction gratings are proposed as an alternative technique to couple a laser diode pump beam into the YAG crystal of a solid-state laser. These binary diffraction gratings are on the long axis of the crystal and are etched in high refractive index coating material. The paper reveals the set of grating parameters and tolerances for transforming vertically incident light into horizontally propagating light inside the crystal with theoretical efficiencies of more than 90%. Under optimal conditions the diffraction grating behaves like an excellent leaky waveguide structure. Theoretical comparisons are made between the efficiencies of gratings directly etched into the laser crystal and gratings etched in to a high index coating material. The resulting zig-zag pumped laser cavity is uniformly excited in order to minimize thermal loads and lensing effects. The maskless binary sub-micron pattern transfer is realized by combining interferometry and lithography.
Image segmentation is one of the major application areas for Pulsed Coupled Neural Networks (PCNN). Previous research has shown that the ability of PCNN to ignore minor variations in intensity and small spatial discontinuities in images is beneficial to image segmentation as well as image smoothing. This paper describes research and development projects in progress in which PCNN is used for the segmentation of three different types of digital images. The software for the diagnosis of Pulmonary Embolism from VQ lung scans uses PCNN in single burst mode for segmenting perfusion and ventilation images. The second project is attempting to detect ischemia by comparing 3D SPECT (Single Photon Emission Computed Tomography) images of heart obtained during stress and rest conditions, respectively. The third application is a space science project which deals with the study of global auroral images obtained from Ultraviolet Imager. The paper also describes an hardware implementation of PCNN as an electro-optical chip.
Pulse Coupled Neural Networks have been extended and modified to suit image segmentation applications. Previous research demonstrated the ability of a PCNN to ignore noisy variations in intensity and small spatial discontinuities in images that prove beneficial to image segmentation and image smoothing. This paper describes four research and development projects that relate to PCNN segmentation - three different signal processing applications and a CMOS integrated circuit implementation. The software for the diagnosis of Pulmonary Embolism from VQ lung scans uses PCNN in single burst mode for segmenting perfusion and ventilation images. The second project is attempting to detect ischemia by comparing 3D SPECT images of the heart obtained during stress and rest conditions, respectively. The third application is a space science project which deals with the study of global aurora images obtained from UV Imager. The paper also describes the hardware implementation of PCNN algorithm as an electro-optical chip.
Diagnoses of cancers and pulmonary embolism are performed by visually interpreting medical data on computer graphics displays. Interpretation aids for medical diagnosis and treatment are not available. The optical information processor system presented in this paper can be used as a second opinion in detecting cancers and classifying images; the final diagnosis is made by a physician. The optical information processing system uses a novel spatial multiplexing technique that allows several images to be processed simultaneously using the same spatial light modulator. Simulation results for liquid crystal display operated in a novel amplitude coupled with binary phase mode is described. In addition, simulation results for a phase modulating micro-mirror spatial light modulator are presented. Results using clinical data show that the optical information processing system can yield a diagnosis rate of 86%.
Pattern recognition algorithms make consistent measureable comparisons among image sets. In this application, normal patient patterns are recongnized and the degree of difference from normal indicates a medical diagnosis of either low- or high-probability of pulmonary embolism. The figure of merit for this study is the vector inner product between the Fourier transforms of each patient image and a filter. The medical application lends itself to implemenation in an optical correlator.
A computer simulation of a two-color holographic interferometric (TCHI) optical system was performed using a physical (wave) optics model. This model accurately simulates propagation through time-varying, 2-D or 3-D concentration and temperature fields as a wave phenomenon. The model calculates wavefront deformations that can be used to generate fringe patterns. This simulation modeled a proposed TriGlycine sulphate TGS flight experiment by propagating through the simplified onion-like refractive index distribution of the growing crystal and calculating the recorded wavefront deformation. The phase of this wavefront was used to generate sample interferograms that map index of refraction variation. Two such fringe patterns, generated at different wavelengths, were used to extract the original temperature and concentration field characteristics within the growth chamber. This proves feasibility for this TCHI crystal growth diagnostic technique. This simulation provides feedback to the experimental design process.
Technological advancements in the field of mixing layer theory have allowed the design and subsequent construction of a Table Top Simulator of Aero-Optic Effects. This experimental facility simulates the supersonic boundary and mixing layers formed by the window coolant gas of an optically guided hypersonic vehicle. This paper discusses the foundations of wave-optic theory applied to model the propagation of optical radiation through such flow. The focus of the calculations is to determine performance quality parameters such as Strehl ratio, jitter, 50 percent contained energy diameter and boresight error. These quality measures will drive the performance requirements of the optical system and focal plane array of the seeker. Comparisons are made between wave-optic model results and actual aero-optic data collected from the Table Top experiment.
Aerodynamic flow surrounding a missile or aircraft in flight can degrade the performance of on-board optical sensor systems.
The minimum resolvable spot or blur circle is a measure of optical system performance. The blur circle size may change by
orders of magnitude throughout the course of the vehicle flight trajectory due to aerodynamic perturbations.
This paper examines the wavelength dependence of blur circle size. It is shown that in many cases an optimum wavelength
exists at which the blur circle size is minimized. Expressions are given for depicting the Point Spread Function shape and
wavelength at which blur is minimized. The optimization expressions presented are suitable for use on a desk top computer