Z. Houkes

Intelligent gas-mixture flow sensor

A simple way to realize a gas-mixture flow sensor is presented. The sensor is capable of measuring two parameters from a gas flow. Both the flow rate and the helium content of a helium-nitrogen gas mixture are measured. The sensor exploits two measurement principles in combination with (local) information handling in an artificial neural network. An analysis of the measurement principles is given and the IC-compatible realization process is described. The sensor is simple to integrate with other micro gas-handling components such as valves, pressure sensors, etc. The sensing elements are combined with a small electronic circuit in which the artificial neural network is implemented. The experimental results are very promising.

Modelling and calibration of the laser beam-scanning triangulation measurement system

We present an approach of modelling and calibration of an active laser beam-scanning triangulation measurement system. The system works with the pattern of two-dimensional beam-scanning illumination and one-dimensional slit-scanning detection with a photo-multiplier tube instead of a CCD camera. By modelling the system-fixed coordinate, we describe the formulation of 3D computation and propose a calibration method in terms of LSE using a planar fitting algorithm. As a sensor-dependent solution, the estimation is refined in the domain of sensing variables. Result of calibration of the real system and a brief analysis of systematic errors are given.

An estimation-based approach for range image segmentation: on the reliability of primitive extraction

This paper presents a new algorithm for estimation-based range image segmentation. Aiming at surface-primitive extraction from range data, we focus on the reliability of the primitive representation in the process of region estimation. We introduce an optimal description of surface primitives, by which the uncertainty of a region estimate is explicitly represented with a covariance matrix. Then the reliability of an estimate is interpreted in terms of “measure of uncertainty”. The segmentation approach follows the region-growing scheme, in which the regions are estimated in an iterative way. With the probabilistic model proposed in this paper, surface homogeneity is defined and tested by an optimal criterion. A notable feature of the algorithm is that the order of merging is organized to lead the growth towards the most reliable representation of the merged region. Concerned with man-made objects in the scene, we restrict the class of surface primitives to be quadric or planar. The proposed algorithm has been applied to real data and synthetic data and demonstrated with experimental results.

A statistical model to describe invariants extracted from a 3-D quadric surface patch and its applications in region-based recognition

A statistical model, describing noise-disturbed invariants extracted from a surface patch of a range image, has been developed and applied to region based pose estimation and classification of 3D quadrics. The Mahalanobis distance, which yields the same results as a Baysian classifier, is used for the classification of the surface patches. The results, compared with the Euclidean distance, appear to be much more reliable.

A new method for fast computation of moments based on 8-neighbor chain code applied to 2-D object recognition

2D moment invariants have been successfully applied in pattern recognition tasks. The main difficulty of using moment invariants is the computational burden. To improve the algorithm of moments computation through an iterative method, an approach for fast computation of moments based on the 8-neighbor chain code is proposed in this paper. Then artificial neural networks are applied for 2D shape recognition with moment invariants. Compared with the method of polygonal approximation, this approach shows higher accuracy in shape representation and faster recognition speed in experiments.

A model driven approach to extract buildings from multi-view aerial imagery

This paper describes a system for analysis of aerial images of urban areas using multiple images from different viewpoints and its evaluation. The proposed approach combines bottom-up and top-down processing. In this paper the emphasis is on the discussion of the experimental evaluation. To evaluate statistically the performance of the system, a set of 100 realisations of 5 images from different viewpoints was used, which was generated by combining real and ray-traced images. The experiments show a significant improvement of reliability and accuracy if multi-view imagery is used instead of single-view.

A study on backpropagation networks for parameter estimation from grey-scale images

A large number of experiments have been done on the basic research of parameter estimation from images with neural networks. To obtain a better estimation accuracy of parameters and to decrease needed storage space and computation time, the architecture of networks, the effective learning rate and momentum, and the selection of training set are investigated. A comparison of network performance to that of the least squares estimator is made. The internal representations in trained networks, i.e. input-to-hidden weight maps or measuring models, which include statistical features of training images and have a clear physical and geometrical meaning, and the internal components of output parameters given by outputs of hidden neurons are presented.<<ETX>>

The estimation of geometry and motion of a surface from image sequences by means of linearization of a parametric

A method is given to estimate the geometry and motion of a moving body surface from image sequences. To this aim a parametric model of the surface is used, in order to reformulate the problem to one of parameter estimation. After linearization of the model standard linear estimation methods can be used to estimate the parameters. The main contribution of this paper is that a method is provided to perform the linearization without specifying the model. Therefore structure from motion estimation and nonrigid body motion estimation can be performed regardless of the model.

Considering shape from shading as an estimation problem

This paper presents a method combining shape and shading models in order to obtain estimations of 3D shape parameters directly from image grey values. The problem is considered as an application of optimal parameter estimation theory, according to Liebelt 8 This theory has been applied previously, where the emphasis was laid on time-delay 2, and motion estimation 3, 5, 9. It is applied here to provide an environment in which somewhat more complicated models can be designed with relative ease and to indicate how the behaviour of the parameters can be investigated. A shading model is added, offering explicit prediction of image grey values. We consider the problem for a single image and for an image pair, showing the shade of the object at two consecutive points of time. The last problem requiresalso a model for the motion of the body. The resulting non-linear estimation problem is linearized about a last parameter guess 8,so that a linear estimator can be applied to compute a new estimate. The various stages of the modelling process are separated by introducing several coordinate systems. Coordinate transformations will show the object from other points of view, and perform an orthographic projection of the 3D scene into the 2D image plane. The explicit grey value prediction yields a template, having a definite extent in the image. Because of the shading model this method requires no gradient images, as in the case of motion estimation 6 or stereo 5. The gradients can be computed analytically. To demonstrate the usefulness and the flexibility of our method, we consider a solid cylinder, irradiated with X-rays. The image is a shadow image originating from the absorption of radiation by the cylinder. In section 2 some background is given about the theory of parameter estimation from digital images. In section 3 the various models for the shape and motion of the body and the imaging process are given. In section 4 and 5 we investigate the properties of the estimator. In section 4 identifiability and uniqueness of the parameters are considered, yielding the parameters, that can be estimated uniquely from the image data. In section 5 some examples are given, elucidating the stabilty properties of the algorithm. To conclude we mention the possibility to replace the motion model with a model connecting images taken from two different positions. Thus this method is also suited to handle a stereo configuration.

A note on conic fitting by the gradient weighted least-squares estimation: refined eigenvector solution

The gradient weighted least-squares criterion is a popular criterion for conic fitting. When the non-linear minimisation problem is solved using the eigenvector method, the minimum is not reached and the resulting solution is an approximation. In this paper, we refine the existing eigenvector method so that the minimisation of the non-linear problem becomes exactly. Consequently we apply the refined algorithm to the re-normalisation approach, by which the new iterative scheme yields to bias-corrected solution but based on the exact minimiser of the cost function. Experimental results show the improved performance of the proposed algorithm.