The problem of active tracking of a ballistic missile during the boost phase is a very challenging task. The Airborne Laser (ABL) program is interested in this problem in that the ABL may use this technique. The Phillips Laboratory in response to this technical requirement embarked on a project to verify the feasibility of active tracking over a horizontal path through the atmosphere. One of the main efforts in this project was to demonstrate active tracking from a ground site, using the Army's Sea Lite Beam Director at the High Energy Laser System Test Facility on White Sands Missile Range. The project proved very successful, both in verifying technical predictions and in gaining hardware experience in accomplishing the active tracking tasks. This paper will present some of the data and review the results of the active tracking missions. The problems encountered included the power of the illuminator beam, the scintillation on the image of the target, noise due to the imaging camera, and jitter in the atmosphere.

The US Air Force Phillips Laboratory is developing the High Altitude Balloon Experiment (HABE) to investigate acquisition, tracking, and pointing concepts to be employed in engagements against boosting missiles in near-space environments. In its most stressing test, HABE employs the Inertial Pseudo Star Reference Unit to provide inertially stabilized line-of-sights (LOSs) for an illuminator laser, active fine track camera, and the marker scoring. The latter serves to measure and score the payload's laser pointing performance. HABE's LOS stabilization subsystem and marker laser pointing are required to demonstrate jitter and drift which is below 1 (mu) rad RMS, a requirement which stresses testing capabilities. At present, a system does not exist to characterize and score the lasers used on this and other experiments at the target plane. This paper will address a concept to provide accurate characterization of laser systems in the far-field target plane.

A real-time executive has been implemented to control a high altitude pointing and tracking experiment. The track and mode controller (TMC) implements a table driven design, in which the track mode logic for a tracking mission is defined within a state transition diagram (STD). THe STD is implemented as a state transition table in the TMC software. Status Events trigger the state transitions in the STD. Each state, as it is entered, causes a number of processes to be activated within the system. As these processes propagate through the system, the status of key processes are monitored by the TMC, allowing further transitions within the STD. This architecture is implemented in real-time, using the vxWorks operating system. VxWorks message queues allow communication of status events from the Event Monitor task to the STD task. Process commands are propagated to the rest of the system processors by means of the SCRAMNet shared memory network. The system mode logic contained in the STD will autonomously sequence in acquisition, tracking and pointing system through an entire engagement sequence, starting with target detection and ending with aimpoint maintenance. Simulation results and lab test results will be presented to verify the mode controller. In addition to implementing the system mode logic with the STD, the TMC can process prerecorded time sequences of commands required during startup operations. It can also process single commands from the system operator. In this paper, the author presents (1) an overview, in which he describes the TMC architecture, the relationship of an end-to-end simulation to the flight software and the laboratory testing environment, (2) implementation details, including information on the vxWorks message queues and the SCRAMNet shared memory network, (3) simulation results and lab test results which verify the mode controller, and (4) plans for the future, specifically as to how this executive will expedite transition to a fully functional system.

The basic kinds of light jumming, which act in flight on the star trackers of the space vehicle attitude control, and the technical solutions used to parry the jumming interference are reviewed in the paper.

Real-time control of gimbaled sensors has been successfully demonstrated using dedicated specialized analog hardware. The reuse of this hardware or its adaptation to new requirements has been limited. To provide more flexibility and allow greater reuse of servo hardware and software, Hughes has developed a commercial off-the-shelf, digitally- based architecture for controlling, measuring and reporting the gimbaled line-of-sight. Applications with multiple servo loops can be performed using real-time hosts and high speed digital signal processors communicating over the VME bus. In this paper we discuss one application of this architecture for a two-axis integrated electro-optical/infrared sensor imbedded in a system with 12 servo loops. The system requirements for gimbal control and line-of-sight reporting required moderately high bandwidth servo loops. Even so, the design incorporated a standard VME bus. Multiple processors with differing interrupt rates were used to control the differing bandwidth servo loops. The requirements, implementation, and performance of this architecture are covered.

This paper presents the application of discrete-point forward-backward Kalman filtering to improve pointing knowledge of the High Resolution Dynamics Limb Sounder (HIRDLS). The instrument is modeled by the following four sections: baseplate, optical bench 1, optical bench 2, and scanner plate. In this system, it is necessary to determine the pointing direction of the scanner with high accuracy. For this purpose, a gyro is affixed to the sounder on optical bench 2. The S/C attitude, extracted from a star- tracker and a second gyro, is also used in this determination. A final piece of information for determining HIRDLS pointing direction is available from the geopotential height at the equator; it has been shown that equatorial geopotential heights are relatively stable, and HIRDLS azimuth scanning is designed to overlap at eh equatorial region so that the same part of the atmosphere will be observed over two consecutive orbits. This condition allows determination of the instrument gyro drift at certain times in each orbit. To take advantage of this information, along with all other data from the various instruments, a discrete-point forward-backward Kalman filter is used to maximize pointing knowledge of the limb sounder. Computer simulation results are provided demonstrating the gains in pointing knowledge.

Science and Technology International (STI) has developed an integrated navigation and stabilization system for the Advanced Airborne Hyperspectral Imaging System (AAHIS). The sensor itself operates as a pushbroom imager, covering the wavelengths range from 435 nm to 830 nm with a ground resolution of 0.2 m² per pixel and a spectral resolution of 182 channels. The system was designed for remote sensing applications utilizing small aircraft. A dGPS navigation system was developed which provides the user with a reliable computer interface for convenient mission planning and monitoring. All data acquisition functions are integrated and require little or no user input. The compact stabilization system consists of two mirrors on a rotation stage mounted in the optical path of the sensor. The navigation and stabilization CPU acquires position data from the differential GPS receiver, as well as attitude data from an inertial navigation system which measures linear and rotational motions. These movements are translated into real-time signals for the mirrors and the rotation stages, correcting for aircraft pitch, yaw and roll and off-track errors. The dGPS data are recorded and later merged with the hyperspectral data which can then be geo-registered and incorporated in a GIS map. The paper will present the approach that was taken to develop a compact stabilization system for hyperspectral imaging in less than 2 MY using off the shelf industrial components. It will quantify the benefit of the chosen approach over other forms of stabilization or no stabilization and discuss the technological and economic benefits for airborne remote sensing missions.

The performance of on Alt-Az telescope depend strongly on its operating conditions. During pointing the telescope can move at a relatively high velocity, and the system can tolerate trajectory position errors higher than during tracking. On the contrary, during tracking Alt-Az telescopes generally move slower but still in a large dynamic range. In this case the position errors must be as close to zero as possible. Furthermore the two pointing and tracking phases are executed in sequence without a well defined switch phase between them. A fixed structure controller cannot optimize the telescope performance in terms of error amount in the whole requested dynamic range. On the contrary a digital controller has the ability to modify its structure and its parameter values during operations, according to the instantaneous error values, system status, system actual speed and system characterization. This paper analyzes the problem of tracking errors and the solutions adopted in a case study.

The Galileo National Telescope is a 3.6 meter Alt-Az telescope installed at the Astronomical Observatory of the Roque de Los Muchachos in La Palma, Canary Islands. The Galileo drive and control systems were designed and developed by the Technology Working Group of the Capodimonte Astronomical Observatory, Naples. This paper reports a description of the systems. The telescope test results in the Ansaldo Company workshop, the present situation and the foreseen improvements of the control system are also described.

The Foxtrot System is an IR simulator that is mounted in a pod under the wing of a test aircraft. It sues a gimbal control system that points and stabilizes an IR camera. The video signal form this camera is sent to an image processor that analyzes the signal in order to detect features of interest in a scene and then make tracking decisions for the gimbal system. The gimbal system has evolved from an analog linear prototype system controlled by an eight bit microprocessor to the present VME bus based system that is controlled by a real time operating system. Starting with the premise that some key systems components are now and expected to remain analog in nature, this paper discusses the evolution of Foxtrot's gimbal system architecture that is taking place in order to accommodate both these required analog interfaces and advances in software and hardware technology. Discussion then focuses on improvements that can be made to the system by mathematically modeling it on a software simulator such as MATLAB. The response of the system can then be described as a difference equation and the emerging capabilities of Digital Signal Processing can be utilized to improve system performance.

A stand-off jammer broadcasting wideband nose causes many problems for a tracking algorithm and is one of the Electronic Counter Measure (ECM) techniques employed in the benchmark problem, Benchmark problems for tracking maneuvering targets have been very helpful in the comparison of proposed algorithms because the problems address some 'real world' tracking issues such ECM, false alarms, and target maneuvers. An accurate estimate of the jammer position and power can be accomplished with passive tracking or through the use of the non-target detections during main- lobe jamming. Recent work in the area of monopulse processing provides a method for debiasing the target measurements in the presence of a jammer and employs dwell averaging through the use of multiple frequencies. Proper beam-pointing is also required to maintain an accurate target track. This paper presents a method for tracking maneuvering targets in the presence of jamming as defined by the benchmark problem. Simulation results that illustrate the performance of the measurement debiasing technique and beam-pointing control are also presented.

A significant problem in tracking and estimation is the consistent transformation of uncertain state estimates between Cartesian and spherical coordinate systems. For example, a radar system generates measurements in its own local spherical coordinate system. In order to combine those measurements with those from other radars, however, a tracking system typically transforms all measurements to a common Cartesian coordinate system. The most common approach is to approximate the transformation through linearization. However, this approximation can lead to biases and inconsistencies, especially when the uncertainties on the measurements are large. A number of approaches have been proposed for using higher order transformation modes, but these approaches have found only limited use due to the often enormous implementation burdens incurred by the need to derive Jacobians and Hessians. This paper expands a method for nonlinear propagation which is described in a companion paper. A discrete set of samples are used to capture the first four moments of the untransformed measurement. The transformation is then applied to each of the samples, and the mean and covariance are calculated from the result. It is shown that the performance of the algorithm is comparable to that of fourth order filters, thus ensuring consistency even when the uncertainty is large. It is not necessary to calculate any derivatives, and the algorithm can be extended to incorporate higher order information. The benefits of this algorithm are illustrated in the contexts of autonomous vehicle navigation and missile tracking.

A multisensor distributed extended Kalman filtering algorithm is presented for nonlinear system, in which the dynamic equation of the system and the equations of sensor's measurements are linearized in the global estimate and global prediction respectively and the suboptimal global estimate based on all available information can be reconstructed from the estimates computed by local sensors based solely on their own local information and transmitted to the data fusion center. An analysis of the properties of the algorithm presented here shows that the global estimate has higher precision than the local one and smaller linearization error than the existing method. Finally, an application of the algorithm to radarflR tracking of a maneuvering target is illustrated. Simulation results show the effectiveness of the algorithm. Keywords: extended Kalman filtering, target tracking, distributed estimation, data fusion

In this paper, the phenomenon of observation multiplicity in the search mode for phased-array radar is studied and a pre- processing algorithm to eliminate redundant observations is proposed. At first, the emergence probabilities of the redundant observations are derived. It is shown that these probabilities depend on signal-to-noise and the spacing between adjacent beams. Second, a priori knowledge of the maximum number of the observations to be merged and the maximum mergence range are introduced to the pre-processing method. At last, the measurement's error before and after the mergence is calculated, and the performance of this method is evaluated and compared with those methods without any limitations.

A difficult problem in multisensor and multi-tracking is that of data association. A multitarget tracking algorithm, probabilistic multi-hypothesis tracking (PMHT), overcomes this problem by estimating the measurement-to-target assignments and the target states simultaneously. We have previously developed two multi-sensor variations of this algorithm, the multi-sensor PMHT and the general multi- sensor PMHT. In this paper, we apply the multi-sensor PMHT algorithm to non-simultaneous radar and optical real data, recorded from a testbed consisting of a radar and optical sensor. Its performance in a multi-target environment is compared to that of a multi-sensor variable update rate Kalman filter.

In a realistic passive sonar environment, the received signal consists of multipath arrivals from closely separated moving targets. The signals are contaminated by spatially correlated noise. The differential MUSIC has been proposed to estimate the DOAs in such a scenario. This method estimates the 'noise subspace' in order to estimate the DOAs. However, the 'noise subspace' estimate has to be updated as and when new data become available. In order to save the computational costs, a new adaptive noise subspace estimation algorithm is proposed in this paper. The salient features of the proposed algorithm are: (1) Noise subspace estimation is done by QR decomposition of the difference matrix which is formed from the data covariance matrix. Thus, as compared to standard eigen-decomposition based methods which require O(N³) computations, the proposed method requires only O(N²) computations. (2) Noise subspace is updated by updating the QR decomposition. (3) The proposed algorithm works in a realistic sonar environment. In the second part of the paper, the estimated bearing values are used to track multiple targets. In order to achieve this, the nonlinear system/linear measurement extended Kalman filtering proposed is applied. Computer simulation results are also presented to support the theory.

Classical methods for multiple target tracking rely on the suboptimal decomposition of the problem into three steps: track initiation, maintenance and deletion. This paper presents an alternative solution that solves the tracking problem in an integrated way, through a global optimization. This problem is known to be a complex combinatorial one, for which neural models are particularly interesting. Previous works, based on the Hopfield model, proposed neural solutions that are drastically sensitive to the choice of some parameters, specially the size of the network and the weights in the cost function to be minimized. This kind of network often converges towards unacceptable solutions. We propose a new neural solution based on a recursive mode, where the constraints are taken into account in introducing competition between the neurons. The network optimizes an objective function that is a measure of the global quality of the tracking, computed as the sum of the track qualities at a given antenna turn. We finally present some simulation results for multiple target monosensor tracking that enable a comparison with classical techniques.

For angle of arrival (AOA), angle resolution and classification of coherent and non-coherent wideband signals will be major problems, especially under the electronic warfare environment. Several methods have been considered for the estimation of the AOA. Multiple signal classification (MUSIC) is one of new suitable methods. But, the method has a disadvantage that it is impossible to estimate the AOA, if the inputs include coherent signal sources such as multipath. In this paper, the array antenna is constructed by some sub-array antennas. The elements of a sub-array antenna are arranged with non-equispace for the classification of noncoherent signals over wideband and some sub-array antennas are also arranged at non-equal distances for the rejection of the angle ambiguity of coherent signals. We applied MUSIC with a spatial smoothing to the array antenna and study how to reject the ambiguity and how to reduce the sidelobe level by using computer simulations. We also make experiments in anechoic chamber to confirm the simulation results. We then indicate that the spatial smoothing of sub-arrays with non-equispace removes the ambiguity of AOA and the classification of coherent signals and noncoherent signals over 8-18GHz is possible.

There are numerous types of real-time processing applications with diverse requirements. Even though application requirements vary significantly, they share many common elements. An integrated image processing software architecture applicable to multiple image processing applications is beneficial in reducing software costs, increasing information fusion, and rapidly prototyping demonstration systems. This paper describes an image processing architecture that provides a framework for integrating multiple image processing applications such as image-based target trackers, IR search and track (IRST), and automatic target cueing/recognition. It discusses the general image processing structure describing common and unique elements of these different applications including such issues as throughput, control and data flow, and latency requirements. We describe the integrated architecture including data input, processing parallelization, image and data processing, information fusing, interfaces, and displays. We present examples of image-based target tracking, moving target indication and IRST implemented using this software architecture on parallel processors. The integrated image processing framework has proven to be extremely beneficial for rapid development of image processing systems form the concept to the demonstration stage.

A hardware-software architecture designed to support image processing applications facilitates rapid development of a fully programmable, high-frame-rate image-based tracker by simplifying and codifying system interfaces. Effective image-based trackers require several key system interfaces for optimal performance, including sensor input, line of sight command output, telemetry storage and/or transmission, and mode and parameter control. Trackers implemented in software are increasingly feasible in many applications because of trends in processor technology, and they are essential as investigatory aids for algorithm development because of their ease of modification. In a software- dominant system, however, interfaces often constitute fully 90 percent of the source lines of code -- every line of algorithm implementation ode requires nine support lines to deal with interfaces. Reducing the support overhead by defining simple, consistent interfaces at the level nearest the tracker software allows developers to create new systems more quickly and change external interfaces with less disruption. We present a fully programmable image-based tracker whose major interfaces were built from a library of reusable components, resulting in a significantly reduced development time. It has supported advanced algorithm research on tracking in both TV and high-frame-rate digital video.

A technique is developed for estimating the leading edge location of an imaged target based on the target's centroid and its slope as the raster is scanned across the target's image. Since the estimate involves more pixels than the standard edge algorithm is easily implemented in a pipe-line process so that the leading edge location can be determined as the video is being read off the focal plane array with minimum processing delay. This algorithm can be shown to be a generalization of the standard biased centroid track algorithm. However, while the biased centroid algorithm requires a priori knowledge of the imaged target shape in order to determine the propose bias, this algorithm estimates the shape dependent factors in real time without any a priori knowledge.

Tracking multiple targets in a cluttered environment is extremely difficult. Traditional approaches use simple techniques to determine what are the true measurements by a combination of gating and some form of a nearest neighbor association. As clutter densities increase, these traditional algorithms fail to perform well. To counter this problem, the multi-hypothesis tracking (MHT) algorithm was developed. This approach enumerates almost every conceivable possible combination of measurements to determine the most likely. This process quickly becomes very complex and requires vast amounts of memory in order to store all of the possible tracks. To avoid this complexity, more sophisticated single hypothesis data association techniques have been developed, such as the probabilistic data association filter (PDAF). These algorithms have enjoyed some success but do not take advantage of any future data to help clarify ambiguous situations. On the other hand, the probabilistic multi-hypothesis tracking (PMHT) algorithm, proposed by Streit and Luginbuhl in 1995, attempts to use the best aspects of the MHT and the PDAF. In the PMHT algorithm, data is processed in batches, thereby using information from before and after each measurement to determine the likelihood of each measurement-to-track association. Furthermore, like the PDAF, it does not attempt to make hard assignments or enumerate all possible combinations. but instead associates each measurement with each track based upon its probability of association. Actual performance and initialization of the PMHT algorithm in the presence of significant clutter has not been adequately researched. This study focuses on the performance of the PMHT algorithm in dense clutter and the initialization thereof. In addition, the effectiveness of measurement attribute data is analyzed, especially as it relates to algorithm initialization. Further, it compares the performance of this algorithm to the nearest neighbor, MHT, and PDAF.

Area correlation tracking is an effective method of tracking targets or a scene of interest that do not have prominent features or high contrast with respect to the background. The essential step in area correlation tracking is to find the 'best fit' of a target/area reference image in a real- time search image acquired by the imaging sensor. The solutions for these are broadly classified as template and feature level matching algorithms. Various algorithms reported in the open literature have different degrees of performance under different field conditions. For certain autonomous tracking applications where there is relative motion between the sensor and the scene/target, it is essential to continuously register a part of previous scene in the next frame. When the reference is thus updated for each frame, there is a possibility that the system may become divergent, if mis-registration is not detected. A correct registration omitted as a wrong registration is tolerable, but a mis-registration committed as a correct registration is a catastrophe. In view of this, it is essential to improve the reliability of registration, and take a confidence measure whenever it is in doubt. This paper addresses that issue and proposes a scheme to improve the performance of area correlation trackers.

This paper deals with the problem of providing missile guidance information by tracking a target object form a sequence of image frames captured from a forward looking IR (FLIR) camera mounted inside a missile moving directly towards the target. Many conventional image based trackers trace key reference points over an image sequence using simple correlation processes. However, the presence of noise and the magnification of image features as the camera approaches the object may cause a tracked point to drift. An alternative technique is to track regions formed by image segmentation. Segmentation is the process of partitioning image pixels into regions of homogeneous intensity levels, for example. However, the difficulty of tracking target regions in a noisy scene, typical of FLIR missile guiding systems, is that regions become divided and therefore there may be little correlation between regions in one frame and the text. This paper introduces a new method that provides a robust solution to the problem of tracking target regions. By acquiring a region template of the target in one frame, the target may be located in a future frame by employing a novel method of performing correlation of the template with a segmented image. For the current design, the target template is acquired manually in the initial frame. In subsequent frames, owing to target magnification, the template is updated automatically after the target has been detected. This new technique produces significant reductions in drift rate compared to a conventional point correlation tracker. Results acquired for real FLIR images obtained from an aircraft are presented.

In this paper, we describe a new method for terminal air-to- ground missile guidance based on IR seeker. The aim is to hit a building which has been previously selected in a 3D model of the scene. The proposed algorithm is divided in two steps: acquisition and tracking steps. Acquisition consists in estimating the location of the target in the first image and to reestimate the missile position. The second step is the tracking of the target along the sequence of images by predicting the target location in each image from the previous one. A supervisor module is in charge of verifying the correctness of the tracking, by doing some reacquisitions in background and ensure the coherence between reacquisitions in background and ensure the coherence between reacquisitions and tracking. All computations are real-time compatible.

In this paper, we derive a filtering method that is based on wavelet packets theory. With the theory we can decompose a length of measurement data into two parts in different resolution levels. One is the low frequency component, the other is the high frequency component. We can detect the target maneuvering respectively in the two components according to different criteria because characteristics of the maneuver information in the two components are not the same. The temporal and spatial location of the target maneuvering can be derived by an 'AND' operation between those of the components. On the other hand, we can also reconstruct the signal with the decomposed components. When in signal reconstruction, we can deliberately filter or even discard some subspaces of the wavelet packets. The reconstruction can be done in certain resolution levels. Thus, by means of subspace selection and reconstruction, we can design a wavelet packets filter with some characteristics, such as the characteristics of lowpass filtering or highpass filtering. With the wavelet packets decomposition and reconstruction, we can not only detect the target maneuver in different resolution level, but we can also attenuate the noise component with little or without attenuation of the target's location and tracking information and maneuver information, providing a 'better' data for the later Kalman filter bank.

This paper presents an approach to detection, tracking, classification and sensor management based on recursive evaluation of a joint multitarget probability. This joint multitarget probability is the conditional probability p_{c(subscript 1}ⁿ,...,c_n(x₁,...,x_n/Z) that there are exactly n targets of class c₁,...,c_n located in cells x₁,...,x_n based on a set of observations Z. This is applied to the problem of estimating the state of a collection of targets moving between discrete cells on a line. A cell can contain more than one target. For the model problem, there are two target classes and the number of targets is not known a priori. The targets are modeled as moving independently with Markov transitions to nearest- neighbor cells. There is one sensor with two modes that can only be used one at a time. These are: a detection mode which can determine whether a cell contains targets but provides no information about target class; and a classification mode that provides little information about the presence or absence of targets but can differentiate between the target classes. For each sensor dwell, the sensor samples a single cell. The conditional probability p_mⁿ,c₁,...,c_n(z/x₁,...,x_n) for sensor output z when mode m is used, given the target location and classes is known. Bayes' rule is applied directly to update the multitarget density for each output. As a basis for sensor management, the expected discrimination gain when a cell is sampled with a particular sensor can be computed. The sensor and cell to maximize the expected gain for each dwell can then be selected. In comparison to directly sampling all of the cells, optimizing the discrimination significantly increases the probability of detecting and localizing the targets.

This paper presents a new concept of tracking systems, called Cinematic Grouping. The partitioning of the observations at different dates between false alarms and tracks is encoded as a matching problem between association hypotheses. This matching problem can be solved efficiently by using metaheuristics like simulated annealing or pulsed neural networks. Some experiments are presented both on synthetic and real scenarii and a comparison with some classical algorithms, like cheap JPDA is performed. It appears that the approach provides very good results, especially in very complex situations, within a reduced computation effort. In fact, the false alarms are well filtered and the tracks are well initialized and maintained.

The accurate debiasing processing for the bias in the classical conversion mainly depends on the first and second statistic of the cosine of the bearing measurement errors. An unbiased conversion is presented. Some comparison between this unbiased conversion with the previously offered debiased conversion is made.

The improved modified gain functions for 3D angles-only tracking are presented, which are more accurate than the originals. Simulation results are presented which demonstrates the superiority of the new modified gain functions.

This paper addresses the problem of estimating the noise equivalent angle (NEA) of several tilt estimation algorithms resulting from including zero-mean Poisson and Gaussian noise. The Poisson noise is due to both photon shot noise and the photoelectric conversion. The readout noise of the detector is assumed to be Gaussian. There are no simple means to relate the noise on the detector to the noise in the measurement. More signal increases the signal-to-noise (SNR) and decreases the NEA. The 2D signal density profile strong influences the NEA. Three different profiles will be analyzed: a Gaussian spot, a top hat, and a simulated missile. The analysis will be performed as a function of total signal. The simulation of these three profiles will be based on using a constant base image with added noise. The variation in the tilt estimate is due to the added noise since each frame would be identical otherwise. The variation as a function of total signal is performed by scaling the base image. Experimental data is analyzed to determine the SNR by dividing the mean by the rms. This data was taken at several different intensity levels so that the total counts would change. Finally, the results from the simulations and the experimental data are compared. The dominant noise not simulated is due to scintillation. Currently, it seems that this last noise source dominates both of the sources included in the simulation.

An unmanned system for automatically 3-axis attitude determination by star constellation matching is a very useful tool for applications primary in the field of satellite technology, where stars are visible all the time. A special designed CCD-camera-system, which is sensitive enough to detect stars, points to the direction to be determinated and captures stars within the camera field of view. By comparing these imaged star pattern with celestial coordinates out of a database-like star catalogue, a star pattern can be matched and the attitude can be calculated by using simple vector algebra. In partnership with other attitude determination systems, the accuracy and reliability for an attitude control system is increasing. Including camera hardware and image processing software, we realized a transputer-based onboard computer system for our own satellite missions. For terrestrial and groundbased pointing applications or post-mission attitude determination a powerful PC-based system is available with additional visualization features of certain steps during the image processing via a user-friendly desktop.

Night sky performance measurements using a star tracker with built in autonomous all stellar attitude estimation software are presented. These measurements validate predictions for on orbit performance. Presented data and statistics include: (1) The time necessary to acquire, track and identify stars and compute an optimal attitude estimate using no a priori attitude knowledge. (2) The star tracker attitude and individual star measurement noise statistics. (3) Star tracking and attitude estimation robustness. Night sky data were collected at the Ball Aerospace facility in Boulder, Colorado on a number of evenings in January 1996. A CT-633 star tracker with its internal star identification and attitude estimation software was pointed at the night sky and rigidly mounted. The CT-633 has a 20 degree field of view and 4.5 magnitude sensitivity. Data were collected at 37 celestial positions in 23 constellations, spanning over 65 percent of the northern celestial hemisphere. Over 250 data sets were collected, all yielding successful star identifications and attitude estimates with no a priori attitude knowledge. In 90 percent of the data sets, stars were acquired, tracked and identified in under one minute with attitude estimates available every 0.2 seconds there after. These data yield CT-633 results consistent with 3 arc-second cross axis attitude estimation performance predicted at beginning of life.

Design practice for the star trackers with the optical system, which consists of a baffle, lens and CCD, shows that necessity to combine optimally the parameters of the mentioned components leads to necessity of taking into account some specific features of their design. It is illustrated by the star tracker whose field of view is equal to 37 degrees.

Necessity to increase and retain the star trackers high accuracy leads to search of the new approaches to their making. The most promising way is providing these devices with a property of self-adapting to the dislocations of their geometric scheme, which consists of the fact that the devices accuracy does not deteriorate at deformation of their geometry. The methods of making nonunbalancing, self- calibrating and combined optical systems of the adaptive star trackers are reviewed, whose usage allows not only provide high accuracy but also reduce a mass of the mechanical components, widen the tolerances at their manufacturing and lower the net cost. The optical systems are presented, that illustrate the given methods.

An autofocus and supervisory positioning system for video- conferencing is showed. The system developed is based in diameter circle detection made by IR pointers. The IR tracks are used after they have been processed as feedforward signals.

There is a growing demand for an automatic surveillance system for road traffic data and industrial workroom environments. These data are required for surveillance and control. The problem of diagnostic intruders in a dangerous areas, knocks generally to the illumination changes. From the beginning of this work, it was stated that, the device had to supervise a robotic environment, in real time, in order to detect the abnormal situations. This paper describes implementation of a fast algorithm of surveillance system that performs tracking of robot's manipulator arm and detection of moving objects. The aim of this work is to avoid collision between human and moving machines. This paper presents a new approach of surveillance allowing unpredictable robotics tasks and tolerant independent illumination changes. We present in our paper an original method to modelize the scene by an image spatial sampling and an algorithm to detect moving objects. The detection is based on the observation of changes between a reference and the current images.

Optical designs for non-scanning, imaging optical system with fields of view exceeding 2p steradians are presented. These optical designs are derived from arthropod visual systems that represents the results of millions of years of evolutionary optical system optimization. Superposition compound eyes, which are the basis for these designs, have evolved on nocturnal insects and deep-water crustaceans. Both of these optically challenging environments are similar to signal poor applications such as missile approach warning where the entire environment of a sensor's platform must be continuously searched for approaching threats. Manmade optical sensor based on superposition compound eyes offer the advantage of no moving parts, no off-axis optical aberrations, no 'dead zones' due to scanning, and a field of view that is limited only by the obscuration of the platform the sensor resides upon. In theory, the maximum field of view of a superposition sensor is 4p steradians, all of which is imaged onto a single image surface. Presented in this paper are designs for manmade versions of these versatile, naturally occurring optical sensors and the results of ray-tracing simulations demonstrating their imaging qualities.