Vital signs, such as heart rate and blood pressure, are critical indicators of patient health and are widely used in clinical monitoring and decision-making. While deep learning models have shown promise in forecasting these signals, their deployment in healthcare remains limited in part because clinicians must be able to trust and interpret model outputs. Without reliable uncertainty quantification - particularly calibrated prediction intervals (PIs) - it is unclear whether a forecasted abnormality constitutes a meaningful warning or merely reflects model noise, hindering clinical decision-making. To address this, we present two methods for deriving PIs from the Reconstruction Uncertainty Estimate (RUE), an uncertainty measure well-suited to vital-sign forecasting due to its sensitivity to data shifts and support for label-free calibration. Our parametric approach assumes that prediction errors and uncertainty estimates follow a Gaussian copula distribution, enabling closed-form PI computation. Our non-parametric approach, based on k-nearest neighbours (KNN), empirically estimates the conditional error distribution using similar validation instances. We evaluate these methods on two large public datasets with minute- and hour-level sampling, representing high- and low-frequency health signals. Experiments demonstrate that the Gaussian copula method consistently outperforms conformal prediction baselines on low-frequency data, while the KNN approach performs best on high-frequency data. These results underscore the clinical promise of RUE-derived PIs for delivering interpretable, uncertainty-aware vital sign forecasts.

Many countries are developing an Urban Air Mobility (UAM) capability defining an Uncrewed Aircraft Systems (UAS) Traffic Management (UTM) architecture to allow safe UAS services in urban environments (e.g., delivery, inspection, air taxis, etc.). The main considerations are air worthiness, operator certification, air traffic management, C2 Link, detect and avoid (DAA), safety management, and security. In addition, if thousands of simultaneous UAS flights are to be achieved, it is not possible for them to be controlled individually by human operators. This makes it necessary to have a rigorous and safe automation methodology to handle such a number of flights. A lane-based airspace structure has been proposed which reduces the complexity of strategic deconfliction by providing UAS agents with a set of pre-defined airway corridors called lanes [1], [2]. This yields collateral benefits including UAS information privacy, robust contingency handling exploiting the lane structure, as well as improved observability and control of the air space. A robust set of UTM parameters and policies must be determined based on the performance characteristics of the deployed UAS platforms, and a methodology which constitutes a first step toward this end is proposed and demonstrated here. In order to realize this approach, a set of initial experiments have been performed to determine the constraints imposed by the UTM on UAS platform capabilities and vice versa. Initial implementation parameters and policies are defined. The major contribution here is a methodology to calibrate UTM safety parameters (e.g., headway, platform speed) in terms of specific platform models' operational characteristics. That is, UTM parameters are a function of platform and not some arbitrarily imposed values. Safety uncertainty is then characterized by the calibration method.

Analyzing patient health through irregular time series vital sign data demands innovative methods beyond conventional imputation techniques. This study introduces a novel approach diverging from prevailing attention-based models to explicitly capture temporal patient evolution. We adopt a paradigm where patients are viewed as dynamic systems evolving over time, with their vital signs encapsulating the system’s states. Our conceptual framework draws parallels to a Markov chain, exploring the transitions between states within a unit of time. To navigate the challenge of a vast state space, we employ a neural network to model expected transitions. Our method portrays the patient’s progression within one unit of time as the system evolves from one state to another, and forecasts states into the future. We outline the training process using irregular time series data and demonstrate its efficacy through analysis on two large vital sign data sets. Comparative analysis against attention-based models emphasizes the effectiveness and efficiency of our approach. This research heralds a promising avenue for patient vital sign analysis, providing insights into temporal patient evolution without relying on imputation methods, thereby enhancing predictive accuracy and interpretability of models.

The FAA and NASA are developing an Advanced Air Mobility (AAM) capability defining an Unmanned Aircraft Systems (UAS) Traffic Management (UTM) architecture. The combined scale and density of the expected air traffic, as well as the algorithmic complexity of maintaining safe separation, are driving a consensus that a structured airspace will eventually be required. Against this background, a lane-based airspace structure is proposed here whose motivation is to reduce the computational complexity of strategic deconfliction by providing UAS agents with a set of pre-defined airway corridors called lanes. To achieve complexity reduction, an airspace is defined that is composed of a directed graph where every node has either input or output degree equal to one, and flight plans consist of a scheduled sequence of lane traversals. The major results are: (1) the creation and layout of lane structures, (2) an efficient lane-based strategic deconfliction scheduling algorithm, (3) lane-network performance analysis tools, and (4) a tactical deconfliction protocol to handle dynamic contingencies (e.g., failure to follow the nominal flight plan). In conclusion, this approach provides efficient scheduling of safe flight paths, straightforward analysis of stream properties of the transportation system, an effective contingency handling protocol, and scalability to thousands of flights over urban areas.

In an effort to meet the changing landscape of education many departments and universities are offering more online courses – a move that is likely to impact every department in some way (Rover et al., 2013). This will require more instructors create online courses, and we describe here how agents and dynamic Bayesian networks can be used to inform this process. Other innovations in instructional strategies are also widely impacting educators (Cutler et al., 2012) including peer instruction, flipped classrooms, problem-based learning, just-in-time teaching, and a variety of active learning strategies. Implementing any of these strategies requires changes to existing courses. We propose ENABLE, a graph-based methodology, to transform a standard linear in-class delivery approach to an on-line, active course delivery system (DuHadway and Henderson, 2015). The overall objectives are: (1) to create a set of methods to analyze the content and structure of existing learning materials that have been used in a synchronous, linearly structured course and provide insight into the nature and relations of the course material and provide alternative ways to organize them, (2) to provide a Bayesian framework to assist in the discovery of causal relations between course learning items and student performance, and (3) to develop some simple artificial student agents and corresponding behavior models to probe the methods’ efficacy and accuracy. In this paper, we focus on our efforts on the third point.

We propose that human generated drawings (including text and graphics) can be represented in terms of actuation processes required to produce them in addition to the visual or geometric properties. The basic theoretical tool is the wreath product introduced by Leyton (Leyton, 2001) (a special form of the semi-direct product from group theory which expresses the action of a control group on a fiber group) which can be used to describe the basic strokes used to form characters and other elements of the drawing. This captures both the geometry (points in the plane) of a shape as well as a generative model (actuation sequences on a kinematic structure). We show that this representation offers several advantages with respect to robust and effective semantic analysis of CAD drawings in terms of classification rates. Document analysis methods have been studied for several decades and much progress has been made; see (Henderson, 2014) for an overview. However, there are many classes of document images which still pose serious problems for effective semantic analysis. Of particular interest here are CAD drawings, and more specifically sets of scanned drawings for which either the electronic CAD no longer exists, or which were produced by hand. We demonstrate results on a set of CAD-generated drawings for automotive parts.

This paper introduces a novel, supportive tool for engineering educators while making course adaptations. As pointed out in the 2013 FIE workshop “An Online Revolution in Learning and Teaching,” online learning is likely to impact every department and teacher in some manner. Other innovations impacting engineering educators include active learning, peer instruction, problem-based learning, and just-in-time teaching. When implementing change, educators are expected to present existing course materials in alternative formats. One resultant difficulty is visualizing, understanding, and judging the impact of various alternatives. Learning materials organization is often limited by delivery methods such as learning management systems that present material linearly. This project uses text analysis and graph transformation techniques to produce various alternatives allowing educators to envision ways changes can be effectively implemented in their courses. We demonstrate how temporal and topical relations between individual learning items can be extracted from existing courses and used to produce a graph that is an effective representation of the course. From this, graph transformations produce alternative organizations of course material allowing various solutions for educators to consider while redesigning their courses. This form of automated brainstorming stimulates out-of-the-box thinking, often producing options previously not considered.

[10] proposed a generative theory of shape, and general cognition, based on group actions on sets as defined by the wreath product. Our position expressed here is that this approach can provide a strong basis for robot cognition when: 1. tightly coupled to sensorimotor data and analysis, 2. used to structure both general concepts and specific instances, and 3. combined with a probabilistic framework (Bayesian networks) to characterize uncer-tainty. We describe a roadmap to achieve these and provide some evidence of feasibility. 1 1

This paper presents a low-complexity, novel approach to wireless sensor network (WSN) assisted autonomous mobile robot (AMR) navigation. The goal is to have an AMR navigate to a target location using only the information inherent to WSNs, i.e., topology of the WSN and received signal strength (RSS) information, while executing an efficient navigation path. Here, the AMR has neither the location information for the WSN, nor any sophisticated ranging equipment for prior mapping. Two schemes are proposed utilizing particle filtering based bearing estimation with RSS values obtained from directional antennas. Real-world experiments demonstrate the effectiveness of the proposed schemes. In the basic node-to-node navigation scheme, the bearing-only particle filtering reduces trajectory length by 11.7% (indoors) and 15% (outdoors), when compared to using raw bearing measurements. The advanced scheme further reduces the trajectory length by 22.8% (indoors) and 19.8% (outdoors), as compared to the basic scheme. The mechanisms exploit the low-cost, low-complexity advantages of the WSNs to provide an effective method for map-less and ranging-less navigation.

In this paper we describe a method for predicting the subjective quality of a new mountain bike route for a particular subject based on routes previously ridden and ranked by the subject. GPS tracks of the previously ridden routes are over laid on rasterized topographic maps and topographic features are extracted in the vicinity of the routes using image processing techniques. The subject ranks each previously ridden route segment on four subjective qualities. The extracted topographic features and the subjective rankings are used as input vectors and target vectors to train a series of decision trees. The decision trees are then tested on a series of route segments not used in the decision tree training. The decision trees were able to exactly predict the subjective rankings with over 60% accuracy vs. 20% accuracy for random selection. When close matches are allowed in the prediction of subjective ranking (plus or minus one point vs. actual) the accuracy of the decision trees increased to 90% and above.

This paper extends the novel research for event localization and target-directed navigation using a deployed wireless sensor network (WSN) [4]. The goal is to have an autonomous mobile robot (AMR) navigate to a target-location by: (i) producing an artificial magnitude distribution within the WSN-covered region, and (ii) having the AMR use the pseudo-gradient from the interpolated distribution in its neighborhood, as it moves towards the target location. Implicit surfaces are used to interpolate the artificial distribution. This scheme only uses the topology of the WSN and received signal strength (RSS) to estimate an efficient navigation path for the AMR. Here, the AMR does not require global coordinates for the region, as it relies on local, neighborhood information alone to navigate. The performance of the scheme is analyzed with hardware experiments and in simulation, using a variety of node-densities and with increasing levels of noise to ensure robustness.

Raster map images (e.g., USGS) provide much information in digital form; however, the color assignments and pixel labels leave many serious ambiguities. A color histogram classification scheme is described, followed by the application of a tensor voting method to classify linear features in the map as well as intersections in linear feature networks. The major result is an excellent segmentation of roads, and road intersections are detected with about 93% recall and 66 % precision.

Many recent deployments of environmental sensor networks have focused on obtaining measurements across large and inhospitable areas. With increasing scale it becomes impractical to deploy or maintain such systems by hand. This paper evaluates large scale network disconnectivity and highlights the underlying issues related to the environment and node characteristics. Furthermore, it examines how a low cost and adaptive method of robotic repair can be applied to large area networks using received signal strength measurements for simple navigation and placement.

We propose computational sensor networks as a methodology to exploit models of physical phenomena in order to better understand the structure of the sensor network. To do so, it is necessary to relate changes in the sensed variables (e.g., temperature) to the aspect of interest in the sensor network (e.g., sensor node position, sensor bias, etc.), and to develop a computational method for its solution. As an example, we describe the use of the heat equation to solve the sensor localization problem. Simulation and physical experiments are described.

Smart Sensor Networks (S-nets) are groups of stationary agents (S-elements) which provide distributed sensing, computation, and communication in an environment. In order to integrate information from individual agents and to efficiently transmit this information to other agents, these devices must be able to create local groups (S-clusters). A leadership protocol that creates static clusters has been previously proposed. Here, we further develop this protocol to allow for dynamic cluster updating. This accommodates on-the-fly network re-organization in response to environmental disturbances or the gain or loss of S-elements. We outline an informal argument for the correctness of this revised protocol. We describe our embedded system implementation of the leadership protocol in simulation and using a colony of robots. Finally, we present results demonstrating both implementations.

We introduced the use of Taring's reaction-diffusion pattern formation to support high-level tasks in smart sensor networks (S-Nets). This has led us to explore various biologically motivated mechanisms. We address some issues that arise in trying to get reliable, efficient patterns in irregular grids with error in inter-node distances.

Sensor networks are comprised of devices having the ability to communicate, compute and sense the environment. A wide range of information processing tasks has been studied for such networks, including operating systems, issues, architecture optimization, and distributed data processing. In this paper, we analyze and compare four different techniques to estimate the gradient of the function represented by the sensor samples. These include: (GA1) a simple device ID defined direction, (GA2) directional derivative, (GA3) polynomial approximation with a plane, and (GA4) polynomial approximation with a quadratic. We compare these based on density of devices per unit area, and noise in the position and sensed data. The interesting result is that GA3 significantly outperforms the other algorithms, although GA1 performs very well and is much easier to compute than the others.

We are engaged in the construction of a smart sensor network to monitor snow conditions and help determine avalanche probability in back country ski areas in the Wasatch mountains. These sensor networks will be comprised of devices having the ability to communicate, compute and sense the environment temperature, light intensity, pressure, and other properties. We have developed several distributed algorithms for such networks and demonstrated them in simulation and small-scale experiments. Our goal is to build and test a 50-100 node network through fall 2004, and to deploy it for experimental study this coming winter.

We have proposed smart sensor networks (S-Nets) as an architecture and set of distributed algorithms to extract, interpret and exploit networked sensor devices. Heretofore, the development of this approach has been done in simulation. In this paper, we describe two complementary implementations of S-Nets: (1) on a set of Berkeley motes comprised of low-power 8-bit, 128 Kb memory processors, communication devices and sensors, and (2) on a set of JStamps having 32-bit controllers, 2 Mb of memory and native execution Java hardware.

This paper describes the development and testing of a new evolutionary robotics research test bed. The test bed consists of a colony of small computationally powerful mobile robots that use evolved neural network controllers and vision based sensors to generate team game-playing behaviors. The vision based sensors function by converting video images into range and object color data. Large evolvable neural network controllers use these sensor data to control mobile robots. The networks require 150 individual input connections to accommodate the processed video sensor data. Using evolutionary computing methods, the neural network based controllers were evolved to play the competitive team game Capture the Flag with teams of mobile robots. Neural controllers were evolved in simulation and transferred to real robots for physical verification. Sensor signals in the simulated environment are formatted to duplicate the processed real video sensor values rather than the raw video images. Robot controllers receive sensor signals and send actuator commands of the same format, whether they are driving physical robots in a real environment or simulated robots agents in an artificial environment. Evolved neural controllers can be transferred directly to the real mobile robots for testing and evaluation. Experimental results generated with this new evolutionary robotics research test bed are presented.

Reverse engineering of mechanical parts requires extraction of information about an instance of a particular part sufficient to replicate the part using appropriate manufacturing techniques. This is important in a wide variety of situations, since functional CAD models are often unavailable or unusable for parts which must be duplicated or modified. Computer vision techniques applied to three-dimensional (3-D) data acquired using noncontact, 3-D position digitizers have the potential for significantly aiding the process. Serious challenges must be overcome, however, if sufficient accuracy is to be obtained and if models produced from sensed data are to be truly useful for manufacturing operations. The paper describes a prototype of a reverse engineering system which uses manufacturing features as geometric primitives. This approach has two advantages over current practice. The resulting models can be directly imported into feature-based CAD systems without loss of the semantics and topological information inherent in feature-based representations. In addition, the feature-based approach facilitates methods capable of producing highly accurate models, even when the original 3-D sensor data has substantial errors.

In previous work (1993,1996), we introduced the notion of instrumented logical sensor systems (ILSS) that are derived from a modeling and design methodology. The instrumented sensor approach is based on a sensori-computational model which defines the components of the sensor system in terms of their functionality, accuracy, robustness and efficiency. This approach provides a uniform specification language to define sensor systems as a composition of smaller, predefined components. From a software engineering standpoint, this addresses the issues of modularity, reusability, and reliability for building complex multisensor systems. In this paper, we demonstrate the practicality of this approach and discuss several design and implementation aspects in the context of mobile robot applications.

We propose to deploy and exploit a large number of inexpensive sensors to obtain information or trigger actions over a wide geographic area. Sensors may be of diverse physical natures: acoustic, IR, seismic, chemical, magnetic, thermal, etc. We describe three major issues: (1) sensor distribution patterns, (2) local sensor frames, and (3) autonomous robot sensor snow exploitation techniques.

Develops theoretical and practical techniques to recover the pose of planar surfaces with minimal sensor readings and motion. A proof is given that two sonar readings suffice to precisely locate a single planar surface that returns the sonar's signal. An implementation is described that takes into account the physical restrictions on actual Polaroid sensors and experiments are described based on actual Polaroid sensor data.

Developing an environment that enables optimal and flexible design of robot manipulators using reconfigurable links, joints, actuators, and sensors is an essential step for efficient robot design and prototyping. Such an environment should have the right "mix" of software and hardware components for designing the physical parts and the controllers, and for the algorithmic control of the robot modules (kinematics, inverse kinematics, dynamics, trajectory planning, analog control and digital computer control). Specifying object-based communications and catalog mechanisms between the software modules, controllers, physical parts, CAD designs, and actuator and sensor components is a necessary step in the prototyping activities. In this paper the authors propose a flexible prototyping environment for robot manipulators with the required subsystems and interfaces between the different components of this environment.

In this work we address the problem of tolerance representation and analysis across the domains of industrial inspection using sensed data, CAD design, and manufacturing. Instead of using geometric primitives in CAD models to define and represent tolerances, we propose the use of stronger methods that are completely based on the manufacturing knowledge for the objects to be inspected. We guide our sensing strategies based on the manufacturing process plans for the parts that are to be inspected and define, compute, and analyze the tolerances of the parts based on the uncertainty in the sensed data along the different toolpaths of the sensed part. We believe that our new approach is the best way to unify tolerances across sensing, CAD, and CAM, as it captures the manufacturing knowledge of the parts to be inspected, as opposed to just CAD geometric representations.

In designing robot manipulators, the interaction between several modules (S/W, VLSI, CAD, CAM, robotics, and control) illustrates an interdisciplinary prototyping environment that includes different types of information that are radically different but combined in a coordinated way. This paper describes the analysis and design of a 3-link robot manipulator prototype as part of a research project for building a prototyping environment for electromechanical systems. This prototype robot can be used as an educational tool in robotics and control classes.>

A perception framework for intelligent inspection and reverse engineering is addressed. The use of discrete event dynamic systems (DEDS) to guide the sensing of mechanical parts is investigated. Dynamic recursive finite state machines (DRFSMs) are introduced as a new DEDS tool for utilizing the recursive nature of the mechanical parts under consideration. The proposed framework uses DRFSM DEDS for constructing an observer for exploration and inspection purposes. A sensing /spl rarr/ CAD interface is constructed for the automatic reconstruction of parts from visual data. A graphical interface for designing DRFSM DEDS controllers is implemented.>

The authors propose an approach to three-dimensional object recognition using computer-aided design (CAD)-based geometry models for freeform surfaces. Geometry is modeled with rational B-splines by defining surface patches and then combining these into a volumetric model of the object. Characteristic features are then extracted and subjected to a battery of tests to select an 'optimal' subset of surface features which are robust with respect to the sensor being used (e.g., laser range finder versus passive stereo) and permit recognition of the object from any viewing position. These features are then organized into a strategy tree that defines the order in which the features are sought, and any corroboration required to justify issuing a hypothesis. The authors propose the use of geometric sensor data integration techniques as a means for formally selecting surface features on free-form objects in order to build recognition strategies.>

The authors explore the connection between CAGD (computer-aided geometric design) and computer vision. A method for the automatic generation of recognition strategies based on the 3-D geometric properties of shape has been devised and implemented. It uses a novel technique to quantify the following properties of features which compose models used in computer vision: robustness, completeness, consistency, cost, and uniqueness. By utilizing this information, the automatic synthesis of a specialized recognition scheme, called a strategy tree, is accomplished. Strategy trees describe, in a systematic and robust manner, the search process used for recognition and localization of particular objects in the given scene. The consist of selected 3-D features which satisfy system constraints and corroborating evidence subtrees which are used in the formation of hypotheses. Verification techniques, used to substantiate or refute these hypotheses are explored. Experiments utilizing 3-D data are presented.>

This paper describes a method for the automatic generation of recognition strategies. This is accomplished using a technique developed for quantifying the following properties of 3-D features which compose models used in 3-D computer vision: robustness, completeness, consistency, cost, and uniqueness. By utilizing this inforniation, the automatic synthesis of a specialized recognition scheme, called a Strategy Tree, is accomplished. Strategy Trees describe, in a systematic and robust manner, the search process used for recognition and localization of particular objects in the given scene. System constraints are satisfied which lead to a set of features which guide the recognition process. Each feature has a Corroborating Evidence Subtrees which validate the initial hypothesis. Verification techniques, used to substantiate or refute these hypotheses, are explored. Experiments are presented.

The split-level relaxation technique is analyzed in a parallel processing framework. It is shown that there is much parallelism inherent in the algorithm that can be exploited. This has been confirmed by implementation of the algorithm on an actual multiprocessor. Although the results are good, the implementation can be made more efficient. The use of multiple queues instead of a centralized queue can reduce memory contention, particularly in large multiprocessors. An asynchronous implementation can also improve the performance.>

Parameter estimation and error analysis techniques applicable to range data are discussed. Extracting surface parameters from noisy range data amounts to data reduction (parameter estimation) and error analysis, the most popular method for which is the least-squares method. Using least squares, it is shown how error in range data percolates upward to manifest itself in the value of each parameter that is calculated either directly or indirectly from data. In addition to estimating surface parameters, errors in the parameters themselves are estimated. The formalisms (with examples) for estimating error in a computed parameter and for combining two or more sets of parameter estimates into a single estimate are presented.>

The authors' goal is the creation of a framework within which independent objectives of a manipulation process can be used to direct a manipulation strategy. The set of contacts that are applied to a task can be partitioned into subsets with independent objectives. A system of this sort is flexible enough to manage large numbers of contacts and to address manipulation tasks that require the removal and replacement of fingers. A simulator has been constructed, and results of its application to position synthesis for initial grasps is presented. The manipulation testbed under construction at the University of Utah using the Utah/MIT dextrous hand are discussed.>

The goal of high level vision is to identify a set of regions in a given image. This has been called by various names: the scene labeling problem’, the consistent labeling problem2, the constraint satisfaction problem3, Waltz filtering4, the satisfying assignment problem5, etc. There are several approaches to solve this problem, including backtracking, graph matching and relaxation. A new method called split-level relaxation, which is based on discrete relaxation was proposed in Ref. 6. It takes care of multiple semantic constraints by considering each of them independently. The problem is known to be NP-complete, so it takes a long time to solve. With the advent of multiprocessors, it is now imperative to see if the problem can be solved faster in the average case. In this paper we give a framework for solving the scene analysis problem in a parallel processing environment, using split-level relaxation. Experiments done on a multiprocessor show that it is indeed advantageous to use multiprocessors to solve this problem.

We describe an approach which facilitates and makes explicit the organization of the knowledge necessary to map robotic system requirements onto an appropriate assembly of algorlthms, processors, sensors, and actuators. In order to achieve this mapping several kinds of knowledge are needed. In this paper, we describe a system under development which exploits the Computer Aided Design (CAD) database in order to synthesize. •recognition Code for vision systems (both 2-D and 3-D), •grasping sites for simple parallel grippers, and •manipulation strategies for dextrous manipulation. We use an object-based approach and give an example application of the system to CAD-based 2-D vision.

Discrete relaxation techniques have proven useful in solving a wide range of problems in digital signal and digital image processing, artificial intelligence, operations research, and machine vision. Much work has been devoted to finding efficient hardware architectures. This paper shows that a conventional hardware design for a Discrete Relaxation Algorithm (DRA) suffers from O(n2m3) time complexity and O(n2m2) space complexity. By reformulating DRA into a parallel computational tree and using a multiple tree-root pipelining scheme, time complexity is reduced to O(nm), while the space complexity is reduced by a factor of 2. For certain relaxation processing, the space complexity can even be decreased to O(nm). Furthermore, a technique for dynamic configuring an architectural wavefront is used which leads to an O(n) time highly concurrent DRA3 architecture.

This paper presents the initial results of our work in using the Computer Aided Geometric Design (CAGD) representations and models as a basis for the visual recognition of 3-D objects for robotic applications. We describe some techniques and algorithms which allow the generation of computer representations and geometric models of complicated realizable 3-D objects in a systematic manner. These representations and models are obtained using (a) available CAGD techniques, and (b) data acquired from various range finding techniques. As compared to previous work in machine vision, multiple hierarchical representations of an object obtained from geometric models can be used for finding orientation and position information.

The representation of 3-D objects is an important step in solving many problems in scene analysis. One of the most successful techniques is that based on the surfaces of objects. We describe several methods for obtaining such surface representations from various types of intrinsic images. In particular, previous work is reviewed and an algorithm based on region growing is investigated in terms of its efficiency in segmenting a set of points in 3-D space into planar faces. Information on the neighborhood structure of the points in the form of a spatial proximity graph is used to direct the segmentation. Applications to industrial objects are demonstrated.

A major application of syntactic pattern recognition is the analysis of two-dimensional shape. This paper describes a new syntactic shape analysis technique which combines the constraint propagation techniques which have been so successful in computer vision with the syntactic representation techniques which have been successfully applied to a wide variety of shape analysis problems. Shapes are modeled by stratified shape grammars. These grammars are designed so that local constraints can be compiled from the grammar describing the appearance of pieces of shape at various levels of description. Applications to the analysis of airplane shapes are presented.