Showing papers on "Object detection published in 1996"

Learning and example selection for object and pattern detection

Abstract: Object and pattern detection is a classical computer vision problem with many potential applications, ranging from automatic target recognition to image-based industrial inspection tasks in assembly lines. While there have been some successful object and pattern detection systems in the past, most such systems handle only specific rigid objects or patterns that can be accurately described by fixed geometric models or pictorial templates. This thesis presents a learning based approach for detecting classes of objects and patterns with variable image appearance but highly predictable image boundaries. Some examples of such object and pattern classes include human faces, aerial views of structured terrain features like volcanoes, localized material defect signatures in industrial parts, certain tissue anomalies in medical images, and instances of a given digit or character, which may be written or printed in many different styles. The thesis consists of two parts. In part one, we introduce our object and pattern detection approach using a concrete human face detection example. The approach first builds a distribution-based model of the target pattern class in an appropriate feature space to describe the target's variable image appearance. It then learns from examples a similarity measure for matching new patterns against the distribution-based target model. We also discuss some pertinent learning issues, including ideas on virtual example generation and example selection. The approach makes few assumptions about the target pattern class and should therefore be fairly general, as long as the target class has predictable image boundaries. We show that this is indeed the case by demonstrating the technique on two other pattern detection/recognition problems. Because our object and pattern detection approach is very much learning-based, how well a system eventually performs depends heavily on the quality of training examples it receives. The second part of this thesis looks at how one can select high quality examples for function approximation learning tasks. Active learning is an area of research that investigates how a learner can intelligently select future training examples to get better approximation results with less data. We propose an active learning formulation for function approximation, and show for three specific approximation function classes, that the active example selection strategy learns its target with fewer data samples than random sampling. Finally, we simplify the original active learning formulation, and show how it leads to a tractable example selection paradigm, suitable for use in many object and pattern detection problems. (Copies available exclusively from MIT Libraries, Rm. 14-0551, Cambridge, MA 02139-4307. Ph. 617-253-5668; Fax 617-253-1690.)

Visual navigation using view-sequenced route representation

Abstract: Previous work in vision-based mobile robotics have lacked models of the route which can be utilized for (1) localization, (2) steering angle determination, and (3) obstacle detection, simultaneously. In this paper, the authors propose a new visual representation of the route, the "view-sequenced route representation (VSRR)." The VSRR is a non-metrical model of the route, which contains a sequence of front view images along a route memorized in the recording run. In the autonomous run, the three types of recognition described above are achieved in real-time by matching between the current view image and the memorized view sequence using a correlation technique. the authors also developed an easy procedure for acquiring VSRRs, and a quick control procedure using VSRRs. VSRRs are especially useful for representing routes in corridors. Results of autonomous navigation using a two-wheeled robot in a real corridor are also presented.

Compact vehicle based rear and side obstacle detection system including multiple antennae

Abstract: A rear and side object detection system for a vehicle (10) based on monolithic millimeter wave integrated circuit technology. A multiple antenna configuration is employed that defines six sensing regions (130,132,134,136,138,140) to the right, left and the rear of the vehicle. A first sensing region (140) is defined at the right side (20) of the vehicle (10), a second sensing region (132) is defined at the right side (20) and rear (22) of the vehicle (10) and overlaps the first sensing region, a third sensing region (136) and a fourth sensing region (130) extend from the rear (22) on both sides (18,20) of the vehicle (10) in the adjacent lanes, a fifth sensing region (134) is defined at the left side (18) of the vehicle and a sixth sensing region (138) is defined at the left side (18) and rear (22) of the vehicle (10) and overlaps the second (132) and fifth (134) sensing regions. A right side warning signal is issued if an object is detected in a right side detection zone defined by the first sensing region or a portion of the second sensing region that does not overlap the third or sixth sensing regions. Likewise, a left side warning signal is issued if an object is detected in a left side detection zone defined by the fifth region and a portion of the sixth sensing region that does not overlap the second or fourth sensing regions. A back-up warning signal is issued if an object is detected in an overlap region (146) between the second (132) and sixth (138) sensing regions.

A unified approach to moving object detection in 2D and 3D scenes

Abstract: The detection of moving objects is important in many tasks. Previous approaches to this problem can be broadly divided into two classes: 2D algorithms which apply when the scene can be approximated by a flat surface and/or when the camera is only undergoing rotations and zooms; and 3D algorithms which work well only when significant depth variations are present in the scene and the camera is translating. In this paper, we describe a unified approach to handling moving object detection in both 2D and 3D scenes, with a strategy to gracefully bridge the gap between those two extremes. Our approach is based on a stratification of the moving object detection problem into scenarios and corresponding techniques which gradually increase in their complexity. Moreover, the computations required for the solution to the problem at one complexity level become the initial processing step for the solution at the next complexity level.

Recognition of planar object classes

Abstract: We present a new framework for recognizing planar object classes, which is based on local feature detectors and a probabilistic model of the spatial arrangement of the features. The allowed object deformations are represented through shape statistics, which are learned from examples. Instances of an object in an image are detected by finding the appropriate features in the correct spatial configuration. The algorithm is robust with respect to partial occlusion, detector false alarms, and missed features. A 94% success rate was achieved for the problem of locating quasi-frontal views of faces in cluttered scenes.

Real-time lane and obstacle detection on the GOLD system

Abstract: This paper describes the GOLD (generic obstacle and lane detection) system, a stereo vision-based hardware and software architecture developed to increment road safety of moving vehicles: it allows detection of both generic obstacles (without constraints on symmetry or shape) and the lane position in a structured environment (with painted lane markings). It has been implemented on the PAPRICA system and works at a rate of 10 Hz.

Assessment of multichannel airborne radar measurements for analysis and design of space-time processing architectures and algorithms

Abstract: System design studies and detailed radar simulations have identified the utility of space-time adaptive processing (STAP) to accomplish target detection in cases where the target Doppler is immersed in sidelobe clutter and jamming. A recent US Air Force investment in STAP has produced a database of multichannel airborne data, through Rome Laboratory's Multichannel Airborne Radar Measurement (MCARM) program, to further develop STAP architectures and algorithms suited to operational environments. An aspect of actual data not typically incorporated into simulation scenarios is the nonhomogeneous features of real-world clutter and interference scenarios. In this paper we investigate the impact of nonhomogeneous data on the performance of STAP. Furthermore, we propose a preliminary scheme to detect and excise nonhomogeneous secondary data in the sample covariance estimation, thereby dramatically improving STAP performance as shown through a specific example using monostatic MCARM data.

Automated detection of human for visual surveillance system

Abstract: This paper describes a robust and reliable method of human detection for visual surveillance systems. The merit of this method is to use simple shape parameters of silhouette patterns to classify humans from other moving objects such as butterflies and autonomous factory vehicles. An extra function based on the brightness level transformation is used to extract the precise shape of the silhouette patterns. An approach to overcome the occlusions of humans is also proposed. We tested our method for 2,500 images (1,100 from humans and 1,400 from other moving objects). Our test system detected the humans at the rate of 98% (=1,077/1,100) and judged 92% (=1,283/1,400) of the other moving objects as non-humans.

Fast stereo based object detection for stop&go traffic

Abstract: This paper presents a new fast binocular stereo approach for the detection and tracking of objects, in particular for stop&go traffic. The proposed scheme is based on a local feature extraction and detects dominant objects by means of a Hough transform like disparity grouping. It is able to handle not only vehicles, but also other objects such as pedestrians. In addition, it allows to determine the camera orientation relative to the road and to distinguish between road surface and obstacles.

3-D object model recovery from 2-D images using structured light

Abstract: Three-dimensional (3-D) object models are currently used in CAD/CAM, robotics, remote sensing, etc. The models (images) can be either directly acquired by using special devices such as range finders, CTR scanners, etc., or they can be recovered from a series of two-dimensional (2-D) images of the object. In this paper, the authors propose a method for determining a set of reference pixels in two simultaneous views of the same object, using two cameras, by projecting a pseudorandom encoded grid on the object. The grid nodes and their encoding values are extracted from 2-D images by applying first a smoothing and then a watershed algorithm. The pseudorandom information encoded in the grid nodes is used to match corresponding sets of points of the two 2-D images. The set of matched points are further used to calculate the disparity of each point of the object surface. Experimental examples illustrate the performance of this simple and elegant technique.

Vector-valued active contours

Abstract: A framework for object segmentation in vector-valued images is presented in this paper. The first scheme proposed is based on geometric active contours moving towards the objects to be detected in the vector-valued image. Objects boundaries are obtained as geodesics or minimal weighted distance curves in a Riemannian space. The metric in this space is given by a definition of edges in vector-valued images. The curve flow corresponding to the proposed active contours holds formal existence, uniqueness, stability, and correctness results. The technique is applicable for example to color and texture images. The scheme automatically handles changes in the deforming curve topology. We conclude the paper presenting an extension of the color active contours which leads to a possible image flow for vector-valued image segmentation. The algorithm is based on moving each one of the image level-sets according to the proposed color active contours. This extension also shows the relation of the color geodesic active contours with a number of partial-differential-equations based image processing algorithms as anisotropic diffusion and shock filters.

Moving object detection apparatus

Abstract: A moving object is detected from a movie which has a complicated background. In order to detect the moving object, there is provided a unit for inputting the movie, a display unit for outputting a processed result, a unit for judging an interval which is predicted to belong to the background as part of a pixel region in the movie, a unit for extracting the moving object and a unit for calculating the moving direction and velocity of the moving object. Even with a complicated background in which not only a change in illumination condition, but also a change in structure occurs, the presence of the structure change of the background can be determined so as to detect and/or extract the moving object in real time. Additionally, the moving direction and velocity of the moving object can be determined.

Genetic Programming for Object Detection

Abstract: This paper examines genetic programming as a machine learning technique in the context of object detection. Object detection is performed on image features and on gray-scale images themselves, with different goals. The generality of the solutions discovered, over the training set and over a wider range of images, is tested in both cases. Using genetic programming as a means of testing the utility of algorithms is also explored. Two programs generated using different features are hierarchically combined, improving the results to 1.4% false negatives on an untrained image, while saving processing.

Method of automatic object detection in image

Abstract: An image is divided into a plurality of subdivision areas. A focus evaluation value representative of the high frequency component contained in the electrical signal from the CCD is calculated for each of the subdivision areas. The object distance is calculated for the respective subdivision areas based on the focus evaluation value. A target object is extracted based on the calculated object distance. The target object is focused. During the focusing operation, when a new object having an object distance smaller than that of the target object is detected in any of remote subdivision areas which have an object distance larger than that of the target object, the target object is re-selected.

Comparison of background extraction based intrusion detection algorithms

Abstract: Several pelwise motion detectors are reviewed in this paper. They are compared in the context of intrusion detection in indoor scenes. The presented motion detection algorithms are based on a pelwise detection of changes in the observed input frame with respect to a recursively updated background. The same global decision module is applied to the outputs of the respective pelwise change detectors. The overall performance is shown to depend on the type of the pelwise temporal filter, and on the image features applied to it.

From active contours to anisotropic diffusion: connections between basic PDE's in image processing

Abstract: We present mathematical and qualitative relations between a number of partial differential equations frequently used in image processing and computer vision. We show for example that classical active contours introduced for object detection by Terzopoulos (1988) and colleagues are connected to anisotropic diffusion flows as those defined by Perona and Malik (1990). We also deal with the relation of these flows with shock filters and variational approaches for image restoration.

Swept-step radar system and detection method using same

Abstract: An apparatus and method for detecting an object and determining the range of the object is disclosed. A transmitter, coupled to an antenna, transmits a frequency-modulated probe signal at each of a number of center frequency intervals or steps. A receiver, coupled to the antenna when operating in a monostatic mode or, alternatively, to a separate antenna when operating in a bistatic mode, receives a return signal from a target object resulting from the probe signal. Magnitude and phase information corresponding to the object are measured and stored in a memory at each of the center frequency steps. The range to the object is determined using the magnitude and phase information stored in the memory. The present invention provides for high-resolution probing and object detection in short-range applications. The present invention has a wide range of applications including high-resolution probing of geophysical surfaces and ground-penetration applications. The invention may also be used to measure the relative permittivity of materials.

Drive assist system using stereo image recognition

Abstract: A safety vehicle using an obstacle detection system with a stereo image sensor has been developed. The stereo image sensor was chosen as a more suitable sensor, compared to a laser radar or a millimeter wave radar. The matching method adopted for this stereo system is the small area based matching to yield a distance distribution image. The road shape and solids are recognized from the distance distribution image. This safety vehicle has three functions for the "avoidance stage" proposed in ASV (advanced safety vehicle) project as follows: a collision alarm system; an autonomous collision avoidance system; and a lane-keeping alarm system. Reduction rates by this safety vehicle for a rear-end collision and a collision encountered at a intersection were estimated. It was recognized that the system was greatly effective for both type of collisions.

Detecting suspicious background changes in video surveillance of busy scenes

Abstract: Detecting background changes in scenes containing significant numbers of moving objects has several applications in video surveillance. One important example is the detection of suspicious packages left in busy airport terminals or train stations. The paper outlines a statistical approach to automatically detecting long term changes to the stationary component of a scene, and describes a prototype system which has been used to successfully demonstrate the feasibility of this approach.

A high-speed algorithm for elliptical object detection

Abstract: A high-speed method for elliptical object location is proposed. Through the use of the geometric symmetry, it first locates the centers of elliptical objects and classifies the boundary points. Then, each point possible on an elliptical object boundary is used to obtain the remaining three parameters of the elliptical object. Since each boundary point is at most transformed to one point on the parameter space, the proposed method is very fast. Some experimental results are also given to show that the proposed method is better than some existing methods under the consideration of space and storage.

Randomized generalized Hough transform for 2-D gray scale object detection

Abstract: This paper proposes a new algorithm for 2D object detection called randomized generalized Hough transform (RGHT). It combines the generalized Hough transform (GHT) with the randomized Hough transform (RHT). Our algorithm can detect arbitrary objects of various scales and orientations in gray level images. We also demonstrate RGHT's advantages of high speed, low storage requirement, high accuracy and arbitrary resolution through comparison with other related algorithms.

Automatic hand-supportable laser scanner and method of reading bar code symbols using same

Abstract: A method and apparatus for reading bar code symbols using a substrate hand-holdable bar code symbol reading device. In general, the automatic bar code symbol reading device comprises a hand-holdable housing containing operative elements which provide an object detection field and a scan field each defined external to the housing. The method involves automatically detecting the presence of an object within the object detection field by sensing object sensing energy reflected off the object. In a preferred embodiment, the object sensing energy is IR radiation produced from an object sensing energy source disposed within the housing. In automatic response to the detection of the object within the object detection field, the hand-holdable device detects the presence of a bar code within the scan field using a laser beam produced within the housing. Then, in automatic response to the detection of a bar code in the scan field, the automatic hand-holdable bar code symbol reading device reads the detected bar code in the scan field by producing scan data signals from the detected bar code and thereafter collecting and analyzing the same. Another aspect of the present invention concerns a hand-holdable data collection device adapted for use with the automatic bar code symbol reading device to form a portable symbol reading system characterized by versatility and simplicity of use.

Method and apparatus for image-based object detection and tracking

Abstract: A method and apparatus for detecting objects from a sequence of images of a scene containing an object by using two distinct methods for object detection. One is suited for well-lit scenes (e.g., daytime), while the other is suitable for poorly-lit scenes (e.g., nighttime) where the objects have lights mounted on them. Further, the method and apparatus are adaptive to the image statistics of the objects being detected, and can be programmed to filter out “weak” detections that lie below a certain percentile in the observed statistical distribution of image measures. The specific percentile is determined based on whether the scene has been determined to be well- or poorly-lit and whether it contains shadows or not.

Automatic detection of interest areas of an image or of a sequence of images

Abstract: The tool introduced in this paper allows to automatically decide in an image or in a video sequence which regions are important and which ones are not. For this purpose, fuzzy logic has been used to modelize human subjective knowledge about the way to allocate priorities to regions. The resulting classification can be used in a wide range of applications going from image coding to image understanding.

Hierarchical MRF modeling for sonar picture segmentation

Abstract: This paper deals with sonar image segmentation based on a hierarchical Markovian modeling. The designed Markov random field (MRF) model takes into account both the phenomenon of speckle noise through Rayleigh's law, and notions of geometry related to the shape of object shadows. We adopt an 8-connexity neighbourhood in order to discriminate geometric and non-regular shadows. MRF are well adapted for this kind of segmentation where a priori knowledge about the shapes we are searching is available. Besides, the introduced hierarchical modeling allows us to successfully improve the sonar image segmentation while speeding up the iterative optimization scheme.

An exposure control system of video cameras based on fuzzy logic using color information

Abstract: Auto exposure (AE) is an important function of video cameras to adjust the image luminance. In this paper, an exposure control system of the AE using color information is discussed. Current AE systems detect special image conditions such as backlighting and excessive frontlighting in which the luminance of a main object deteriorates, and compensate the exposure in order to obtain the appropriate luminance of the main object. The compensation is determined according to the degree of backlighting and excessive frontlighting, regardless of the the background. The exposure control system proposed in this paper uses "hue" and "chroma" of pixels to derive the importance of the background, and determines the amount of compensation required by fuzzy reasoning. Simulations of AE were carried out using both the conventional system and proposed method. Results showed that the proposed system was more efficient for AE than the conventional method.

Detection and location of moving objects using deterministic relaxation algorithms

Abstract: Two important problems in motion analysis are addressed in this paper: change detection and moving object location. For the first problem, the inter-frame difference is modelized by a mixture of Laplacian distributions, a Gibbs random field is used for describing the label field, and HCF (highest confidence first) algorithm is used for solving the resulting optimization problem. The solution of the second problem is based on the observation of two successive frames alone. Using the results of change detection an adaptive statistical model for the couple of image intensities is identified. Then the labeling problem is solved using HCF algorithm. Results on real image sequences illustrate the efficiency of the proposed method.

Object detection apparatus

Abstract: The presence/absence of an object, movement of the object, the direction of movement, the speed of movement, the shape of the object, the quantity of objects and time information related to these factors are detected without installing a means for transmission An antenna device (1) has a function of receiving radio waves (EW1 to EW24) transmitted by GPS satellites and outputs a detection signal when the reception of the radio waves (EW1 to EW24) transmitted by the GPS satellites is blocked by an object (A) A signal processing device (2) detects the object (A) based upon the detection signal provided by the antenna device (1)

Evaluation of image metrics for target discrimination using psychophysical experiments

Abstract: Image clutter affects the perceptual ability of any system for object detection. A procedure for conducting psychophysical experiments has been developed to test computational models for the perceptual similarity or difference of texture patterns, which contributes to image clutter. This experimental procedure is based on Thurstone’s law of comparative judgment, which is used along with the method of paired comparisons to assign relative psychological scale values to image stimuli. To facilitate consistency in the presentation of stimuli and collection of data, an X-windows testing environment has been developed called the X-based perceptual experiment testbed. Using this experimental procedure, a pilot study was conducted in which the image stimuli consisted of targets and backgrounds with texture patterns of uncorrelated Gaussian noise. With such patterns, only first-order image statistics are of significance. The psychological scale values relating the level of ‘‘target distinctness’’ in each of the image stimuli were compared to several first-order image metrics. Correlation coefficients as high as 0.9881 were found between the scale values and the image metrics.