Showing papers on "Object detection published in 1989"

Detecting moving objects

Abstract: The detection of moving objects is important in many tasks. This paper examines moving object detection based primarily on optical flow. We conclude that in realistic situations, detection using visual information alone is quite difficult, particularly when the camera may also be moving. The availability of additional information about camera motion and/or scene structure greatly simplifies the problem. Two general classes of techniques are examined. The first is based upon the motion epipolar constraint—translational motion produces a flow field radially expanding from a “focus of expansion” (FOE). Epipolar methods depend on knowing at least partial information about camera translation and/or rotation. The second class of methods is based on comparison of observed optical flow with other information about depth, for example from stereo vision. Examples of several of these techniques are presented.

Object detection and classification using matched filtering and higher-order statistics

Abstract: A method for signal detection and classification in the presence of additive Gaussian noise using higher-than-second-order statistics of the matched filter output is presented. Deterministic and random, nonGaussian distributed signals are detected via multiple correlations and cumulants, respectively. The detection algorithm is computationally simple, and, contrary to standard matched filtering, it is insensitive to signal shifts and does not require knowledge of the noise spectrum for prewhitening. The detector can be viewed as a likelihood radio test between sampled higher-order statistics, and its performance is evaluated using binary hypothesis testing. Signals are designed to have equal higher-order correlation energies and then classified based on higher-order statistics. Two-dimensional extensions of the one-dimensional algorithms are discussed briefly. Simulations illustrate successful performance of the detection and classification algorithms at low signal-to-noise ratio. >

Object detection by tactile sensing method employing force/torque information

Abstract: An object shape detection system using a force/torque sensor and an insensitive probe is proposed. the development of a tactile sensing method is which makes it possible to estimate a contact position on the probe by measuring only contact force is presented. Though the probe itself is not sensitive, the method derives the contact position from force information measured by the sensor. The estimated values do not depend on either the magnitude or the direction of the force. The method can be with any probes of any shape and material. This method is applied to a manipulator system, and object shape detection experiments are conducted using two types of probes. Shapes of three-dimensional objects are successfully reproduced. The results confirm that the method enables the manipulator system to detect objects without precise control. >

Biological object detection apparatus

Abstract: An apparatus for detecting and identifying a biological object. A transparent plate has a first surface onto which a light beam is projected and a second surface onto which a biological object to be detected and identified is placed. The light beam projected toward the first plate surface is transmitted through the plate and toward the object on the second surface, from which the light beam is reflected and retransmitted through the plate toward and through the first surface thereof and received and detected by an optical detector. The detection of a biological object is confirmed by comparing the change of the wavelength characteristics of the reflected and detected light beam in a predetermined time sequence according to the object being first placed upon and then pressed upon the second surface of the transparent plate with respective, known such characteristics thereof.

Knowledge-based approach to sleep EEG analysis-a feasibility study

Abstract: A knowledge-based approach to automated sleep EEG (electroencephalogram) analysis is described. In this system, an object-oriented approach is followed in which specific waveforms and sleep stages (objects) are represented in terms of frames. The latter capture the morphological and spatiotemporal information for each object. An object detection module (frame matcher), operating on the frames, is used to identify what features used to be extracted from the EEG and to trigger the appropriate specialist-specialized signal processing modules-to obtain values for these features. This leads to an opportunistic approach to EEG interpretation with quantitative information theory being extracted from the signal only when needed by the reasoning processes. The system has been tested on the detection of K complexes and sleep spindles. Its performance indicates that the approach is feasible. >

Image sequence compression by iconic indexing

Abstract: A method that facilitates the rapid retrieval of a given image sequence from a large database is presented. The authors exploit the fact that much of the information stored is redundant. They extend the two-dimensional string methodology to image sequences. This permits queries on the relative positions of objects within video sequences, including changes in position over time. >

Determination of three-dimensional object location and orientation from range images

Abstract: A technique for determining the distortion parameters (location and orientation) of general three-dimensional objects from a single range image view is introduced. The technique is based on an extension of the straight-line Hough transform to three-dimensional space. It is very efficient and robust, since the dimensionality of the feature space is low and since it uses range images directly (with no preprocessing such as segmentation and edge or gradient detection). Because the feature space separates the translation and rotation effects, a hierarchical algorithm to detect object rotation and translation is possible. The new Hough space can also be used as a feature space for discriminating among three-dimensional objects. >

Constraints for the computation of optical flow

Abstract: A number of constraints are proposed for which both the components of optical flow can be obtained by local differential techniques and the aperture problem can usually be solved. The constraints are suggested by the observation that it is possible to describe spatial and temporal changes of the image brightness in terms of infinitesimal deformations. An arbitrary choice of two of the four equations which correspond to the elementary deformations of a 2-D pattern implies that the spatial gradient of the image brightness is stationary and leads to a linear system of equations for optical flow which seems best suited for numerical implementation on real data in the absence of a priori information. In that case, the error term between the computed optical flow and the motion field-that is, the 2-D vector field associated with the true displacement of points on the image plane-is derived and the conditions under which it can safely be neglected are discussed. Experiments on real images are reported which show that the obtained optical flows allow the estimate of 3-D motion parameters, the detection of discontinuities in the flow field, and the segmentation of the image in different moving objects. >

Object detection from a real scene using the correlation peak coordinates of multiple circular harmonic filters

Abstract: Most proposed methods for rotation invariant multiple circular harmonic filters combine the intensities or amplitudes of the center correlations. We propose combining the coordinates of the correlation peaks of multiple circular harmonic filters. The new method is used to detect objects from a cluttered noisy input scene. High cross-correlation peaks coming from clutter do not cause false alarms. Experimental results are shown.

Motion detection in an image sequence using Gibbs distributions

Abstract: The authors address the problem of motion detection in an image sequence from the variations in time of the intensity distribution. The problem is not limited to change detection but encompasses the recovery of the projections of moving areas in the image. The approach is characterized by the joint treatment of the detection of temporal changes and the reconstruction of mobile object masks according to a probabilistic formulation. More formally, spatio-temporal contextual information is introduced through Markovian models, using Gibbs distributions defined on a spatio-temporal neighborhood system. Then the problem at hand is stated as a statistical labeling one. To decide whether or not a point belongs to a moving area is equivalent to assigning to it a given label. A solution to this labeling problem is formulated according to the maximum a posteriori (MAP) criterion. Experiments with a real image sequence have been carried out. >

Side Scan Sonar Object Classification Algorithms

Abstract: Autonomous underwater vehicles require the capability to understand their environment. This understanding, coupled with the operational goals of the vehicle, determines the subsequent actions of the vehicle. Environmental understanding is realized through the vehicle's sensors and a priori knowledge. This paper focuses on our investigation of automatic interpretation of side scan sonar data for the purpose of detecting and classifying undersea mines. The test data set is a series of side scan sonar images taken from a U. S. Navy acoustic sensor under optimal conditions (flat, sandy bottom). Groundtruth is available for acoustic images with eight unique types of mine targets. The interpretation of the data is per formed in two stages. The first stage, preprocessing and target detection, uses an adaptive thresholding algorithm coupled with an adaptive averaging technique to locate objects of interest in the sonar image. The second stage, classification, performs a binary classification of whether each detected object is, or is not, a mine. The classification is achieved using an attribute-based decision tree. An approach for a third stage, identification of the mark and mod of the classified bottom mines, is also presented. The results to date, as well as plans for the future, are discussed.

Automatic Target Detection and Cuing System for an Autonomous Underwater Vehicle (auv)

Abstract: This paper presents the technology needed for searching large ocean areas with imaging sonars or video cameras. To search a large area in reasonable time, multiple autonomous underwater vehicles (AUVs) may be used to cover contiguous parts of the search area. This autonomous search capability allows operating personnel to be removed from hazardous environments, provides economy through reduction in the number of required host ships, and reduces search time. The technical approach presented here uses both algorithms and neural networks to build a target detection and classification system. This system is robust in that it continues correct operation even with multiple failures. Hardware implementations of the system allow processing at video and ultra high resolution sonar rates. Both sonar and video images are created from either side scan sonar or video camera using 8 bit grey level pixel representations of the reflected energy. Objects are detected by searching for highlight, shadow, texture changes or statistical anomalies in the image. A neural network is also trained to recognize image features and it is used to complement the performance of the algorithmic detectors. A neural network (NN) has been developed for classifying target scenes which are defined as specific sets of man-made and natural features. The NN classifier has been trained to recognize the target scene from 360 degree aspect angles. Detector and classifier performance are evaluated using both our synthetic and real image libraries. We present our current laboratory hardware for testing and training both the algorithms and neural networks. Hardware for real time execution is also discussed.

Top-down image segmentation using object detection and contour relaxation

Abstract: A novel segmentation technique that starts with the whole image being a single region is presented. First, an object detection scheme, which marks those locations where local statistics deviate significantly from the overall statistics, provides location and approximate shapes of the major objects (regions) in the scene. Exact boundaries are subsequently obtained by a contour relaxation algorithm, which includes a general model for typical region shapes. Object detection and contour relaxation are repeated recursively until a stable segmentation result is achieved. Segmentation results are presented. >

Object detection using two channel active optical sensors

Abstract: An apparatus for detecting target objects which exhibit ratios of reflectance values that are substantially the same at two separated wavelengths when compared with background objects having reflectance values that are substantially different at the same two wavelengths has a two channel active optical sensor and logic circuitry. A first signal is transmitted at a wavelength lambda-1 which is different from a second signal which is transmitted at a wavelength lambda-2. The two signals are directed towards an object which reflects the signals therefrom. The reflected signals are received and a ratio is calculated. The ratio is compared with a predetermined threshold value to indicate when a target object has been identified.

Efficient Detection of Small Moving Objects

Abstract: : A signal processing problem encountered with many sensor systems having a wide field-of-view is detection of small, unresolved objects moving in a straight line amid stationary clutter. The wide field-of-view combined with the need to accurately pinpoint object positions imply that these sensors must have hundreds of thousands of samples in their output. To process this amount of data in a timely fashion, computationally efficient algorithms are a necessity. In this report, a computationally efficient set of algorithms is described for detecting satellites, meteorites, and other moving objects using data from an optical telescope charge-coupled device (CCD) focal plane in the MIT Lincoln Laboratory Demonstration Surveillance System (DSS). The trade-off of reduced detection sensitivity for lower computational cost in the algorithm is quantitatively discussed. Major techniques employed are: 1. Sample normalization by temporal mean and standard deviation to suppress clutter, 2. Maximum value projection to reduce the dimensionality of the data; 3. A two-stage matched filter detector which first nominates and then confirms signal candidates, and 4. Two-dimensional binary velocity filtering. The techniques should have practical application to other wide field-of-view sensors where moving object detection is important.

Focusing linear arrays operating in a sequential mode

Abstract: The authors describe multiple target detection using a linear array, operated in the sequential mode. In this approach, the firing of the elements is sequential: one element is transmitting with all the elements receiving. A novel processing technique is presented which allows multiple target detection without tapering or phase shifting hardware, and which includes focusing at any range. Three examples are presented for a real array in the sequential mode: (1) detection of far field scatterers with focusing at infinity, (2) detection of finite range scatterers with focusing at infinity, and (3) detection of finite-range scatterers with proper focusing. The sequential operation and processing allows the simultaneous detection of all the targets within the elementary 3-dB beam. The synthesized array pattern is shown to be much better than that of the standard parallel operation, which requires a set of phase shifters of each direction sine. It is concluded that the implementation of sequential processing is straightforward and storage is minimized by accumulating the data in terms of phase index. >

A New Collision Detection Algorithm Using Octree Models

Abstract: The octree representation has been recently proposed as a tool for collision detection purposes. Nevertheless, the applications described only involve using an octree to model the -environment, but not the robot or other moving bodies. This situation is due to the relatively high cost of the octree rotation algorithms. In this paper, we present a method based on the simultaneous traversal of the environment tree and the octree of the mobile. This technique allows considerable savings in the number of nodes computed and leads to an efficient method for collision detection.

Multiresolution object detection and segmentation using top-down algorithms

Abstract: Multiresolution approaches to image segmentation are being investigated widely as potential platforms on which to perform real-time operations on video sequences. A pyramid is a massively parallel computational platform on which variable resolution representations of an underlying image are used to obtain a segmentation. The paper is concerned with the use of computationally efficient hierarchical techniques for object detection and segmentation, and describes several such algorithms which exploit the pyramid structure using vertical interactions between levels. The algorithms use top-down approaches to achieve good performance at a lower cost relative to iterative techniques. The algorithms are discussed and their performance on both synthetic images and real infrared images is compared in terms of segmentation quality and computational cost. Results using iterative linking procedures are also presented, and are compared with the present algorithms in terms of cost and performance. The segmentation performance of the top-down algorithms is shown to be comparable to that of the more computationally expensive iterative algorithms.

Object Detection and Analysis in Digitised Images: Algorithmic Components and Software Structure

Abstract: The structures of major 2-dimensional image analysis software packages in optical astronomy are overviewed. Reference is made to the varied image analysis and pattern recognition algorithms in use.

3D object recognition via simulated particles diffusion

Abstract: A novel approach for 3D object recognition is presented. This approach is model-based, and assumes either 3D or 2/sup 1///sub 2/D scene acquisition. Transformation detection is accomplished along with an object identification (six degrees of freedom, three rotational and three translational, are assumed). The diffusion-like simulation recently introduced as a means for characterization of shape is used in the extraction of point features. The point features represent regions on the object's surface that are extreme in curvature (i.e. concavities and convexities). Object matching is carried out by examining the correspondence between the object's set of point features and the model's set of point features, using an alignment strategy. Triangles are constructed between all possible triples of object's point features, and then are aligned to candidate corresponding triangles of the model's point features. 2/sup 1///sub 2/ range images are transformed into a volumetric representation through a parallel projection onto the 3-D space. The resultant volume is suitable for processing by the diffusion-like simulation. >

Parallel Implementation Of The Split And Merge Algorithm On Hypercube Processors For Object Detection And Recognition

Abstract: Split and merge is a computationaly efficient region segmentation technique suitable to detect objects or surfaces in a given image Despite its superior performance, it suffers from large memory usage and excessive computation time This paper describes parallel implementation of the split and merge algorithm in a 16 node hypercube processor in order to reduce processing time to an acceptable level in the real time applications Three methods are proposed to parallelize the operation of the method using the nearest neghbor (mesh) topology that can be mapped onto the hypercube architecture Comparison of the described techniques is given and processing results of the real world images are presented

Orientation measurement of piece parts using single-layer networks

Abstract: A technique for measuring the orientation of elongated planar objects using a number of single-layer networks is presented. This technique has been developed to be used in a noisy multiobject environment and can cope with object registration. The network consists of five discriminators, each of which has the capability to detect the directionality of an object in a certain range of angles. The memory elements of each discriminator are divided into groups. The memory elements within each group are summed and thresholded, and the outputs of the various groups within each discriminator are summed to form the output coming from the various discriminators. The connection of the memory elements is arranged in this way in order to detect the direction of the rectangle which confines the object. The network can measure the orientation of an object in the range of -10 degrees to +90 degrees , but this is because only five discriminators were used in the present experiment. >

Noise Measurement And Suppression In Active 3-D Laser Based Imaging Systems

Abstract: Active, laser based 3-D sensing techniques can provide several practical advantages for direct, real time depth measurement. Numerous active techniques exist ranging from camera based structured light used for object detection and height measurement through scanning laser radar systems targeted for complex tasks like robot navigation. In all systems the performance is limited by the level of optical and electronic noise in the system. This paper identifies the fundamental sources of optical and electronic noise and presents data which can be used to estimate the performance of several system configurations. The data is generally useful for improving the signal detection capability of many types of imaging systems. Implementation of noise reduction techniques can provide extraordinary optical and electronic dynamic range, particularly when laser scanning 3-D imaging systems are used.

Optical switching devices

Abstract: An optical switching device for detecting a remote object comprises an oscillator 120, a light-emitting circuit 130 for emitting towards the remote object an infra-red light pulse train corresponding to a pulse train from the oscillator 120, a light-receiving circuit 140 for converting a light pulse train reflected from the remote object into an electrical detection pulse train, a synchronous circuit 150 for synchronizing the frequency of the detection pulse train with a reference signal of adjustable frequency to provide a logic signal when the detection pulse train and reference signal are in synchronism, and a switching circuit 160 for generating object detection logic signals in response to the logic signal from the synchronous circuit. The synchronous circuit 150 may include a phase-locked-loop circuit. In modifications the light pulse train may be transmitted through the object, or the object may carry the oscillator 120 and light-emitting circuit 130.

Object detection and recognition via stochastic model-based image segmentation

Abstract: Summary form only given. A stochastic model-based image segmentation technique that utilizes the tone descriptor for object detection and recognition has been developed. The image regions are characterized by region-dependent constant mean (average-gray level) and variance (variation of gray level), and the distribution of the regions is modeled by a stochastic model. For a nondiffracting computed tomography (CT) image it has been proved that (1) at the pixel level, the pixel images are the asymptotic normal random variables, (2) at the class level, the regions are a normal random field, and (3) at the picture level, the observed image is a finite normal mixture. >

Exploiting temporal coherence in scene analysis for autonomous navigation

Abstract: A technique for building reliable scene descriptions by evaluating the temporal stability of detected objects is presented. The approach is designed to avoid mistakes and to increase the competence of the sensing system by tracking objects from image to image and evaluating the stability of their descriptions over time. Since the information available about an object can change significantly over time, the authors introduce the idea of a representation space, which is a lattice of representations progressing from crude blob descriptions to complete semantic models, such as bush, rock, and tree. One of these representations is associated with an object only after the object has been described multiple times in the representation and the parameters of the representation are stable in a statistical sense enhanced by a set of explanations describing valid reasons for deviations. To illustrate the power of these ideas, the authors have implemented a system, called TraX, that constructs and refines models of outdoor objects detected in sequences of range data. >

Advances in image compression and automatic target recognition; Proceedings of the Meeting, Orlando, FL, Mar. 30, 31, 1989

Abstract: Various papers on image compression and automatic target recognition are presented. Individual topics addressed include: target cluster detection in cluttered SAR imagery, model-based target recognition using laser radar imagery, Smart Sensor front-end processor for feature extraction of images, object attitude estimation and tracking from a single video sensor, symmetry detection in human vision, analysis of high resolution aerial images for object detection, obscured object recognition for an ATR application, neural networks for adaptive shape tracking, statistical mechanics and pattern recognition, detection of cylinders in aerial range images, moving object tracking using local windows, new transform method for image data compression, quad-tree product vector quantization of images, predictive trellis encoding of imagery, reduced generalized chain code for contour description, compact architecture for a real-time vision system, use of human visibility functions in segmentation coding, color texture analysis and synthesis using Gibbs random fields.