Rotation Invariant Pattern Recognition Research Articles

Scale and rotation invariant pattern recognition using complex-log mapping and augmented second order neural network

A scale and rotation invariant pattern recognition system using complex-log mapping (CLM) and an augmented second order neural network (SONN) is proposed. CLM is very useful for extracting the scale and rotation invariant features. The results are, however, given in a wrap-around translated form. This problem is solved with an augmented SONN. Experimental results show that the proposed system has improved recognition performance.

Circular harmonic filters for a rotation-invariant incoherent correlator

The incoherent optical correlator was extended to perform rotation-invariant pattern recognition by using a set of two real-positive filters presenting a single-order circular harmonic function.

Multichannel rotation-invariant pattern recognition for polychromatic objects using circular harmonic filters

A method for rotation invariant pattern recognition of polychromatic objects is studied. The recognition of polychromatic objects is obtained by using a multichannel technique. The method is based in the obtention of three correlations for three colour channels (red, green and blue). The rotation invariance is achieved by using circular harmonic filters in each channel.

A method for shift, rotation, and scale invariant pattern recognition using the form and arrangement of pattern-specific features

A new computer based method for visual pattern recognition is described. The process extracts shape features from binary edge images represented in either real-valued or discrete coordinates. Both separate and overlapping features are extracted, maximizing their completeness and continuity, without using templates or mathematical feature descriptions. These features and their arrangement are encoded in a shift, rotation, and scale invariant manner and used to recognize incoming patterns from a set of previously encoded patterns stored in memory. New patterns may be automatically classified and stored in memory, allowing the system to learn and generalize without operator intervention. Testing shows the system to be completely shift, rotation, and scale invariant, and insensitive to local image distortion.

Translation, rotation, and scale invariant pattern recognition by high-order neural networks and moment classifiers

The classification and recognition of two-dimensional patterns independently of their position, orientation, and size by using high-order networks are discussed. A method is introduced for reducing and controlling the number of weights of a third-order network used for invariant pattern recognition. The method leads to economical networks that exhibit high recognition rates for translated, rotated, and scaled, as well as locally distorted, patterns. The performance of these networks at recognizing types and handwritten numerals independently of their position, size, and orientation is compared with and found superior to the performance of a layered feedforward network to which image features extracted by the method of moments are presented as input.

Rotational invariant pattern recognition with the method of circular harmonics using a BaTiO 3 crystal

We have implemented the method of circular harmonic (CH) filtering in a photorefractive crystal to perform pattern recognition via four-wave mixing. Experimental results have shown that the recognition result is fully rotational-and shift-invariant. The crystal we have used is BaTiO 3.

Invariant pattern recognition using Fourier-Mellin transforms and neural networks

The authors present image descriptors and spatial filters based on the circular Fourier and radial Mellin transforms. The Fourier-Mellin descriptors have advantage in terms of the noise sensitivity. The Fourier-Mellin filters generate the shift invariant Fourier-Mellin transforms optically. An opto-neural system using the Fourier-Mellin filters and a back-propagation neural network shows shift, scale and rotation invariant pattern recognition.

Circular harmonic covariance filters for rotation invariant object recognition and discrimination

We propose using the covariance function as a measure of similarity in the circular harmonic matched filters for rotation invariant pattern recognition and for discrimination against unwanted objects which are unknown a priori. An optical covariance filter is introduced, which is a modified circular harmonic composite filter combining a conventional circular harmonic filter with a uniform circular harmonic function. Both optical and computer experiments show improved discrimination capability.

Maximum discrimination filter

Spatial filters for discrimination among the members of a training set are shown to be available with a global optimum. These filters can be implemented with real bipolar point-spread functions. The method is demonstrated to be effective also for rotation-invariant pattern recognition.

Incoherent pattern recognition with limited rotation invariance

The application of circular cosine expansion to intensity pattern recognition is investigated. Choosing one modified component of the circular cosine expansion, a positive reference pattern is composed and limited rotation-invariant incoherence pattern recognition is achieved. Computer simulation shows that it is in good agreement with the theoretical prediction.

Recording holograms with diagonal coding on binary spatial light modulators for pattern recognition

The performance of computer-generated holographic matched filters recorded on 128 x 128 pixel supports was studied experimentally, with the objective of recording holograms on spatial light modulators such as the Semetex SIGHT-MOD for rotation-invariant pattern recognition. A new method of diagonal coding is introduced. This method yields good results, as it allows a more accurate phase quantization. Experimental results from computer simulations show that high speed rotation-invariant recognition of simple shapes is possible with such holograms.

Rotation invariant pattern recognition with a programmable joint transform correlator

A circular harmonic expansion filter is real value implemented in a joint transform correlatora rchitecturet o perform rotation invariant pattern recognition.

Rotation-invariant digital pattern recognition using circular harmonic expansion: author’s reply to comments

As Danielsson says in his comment on the paper by Arsenault, the phase information is not used and the same signature vectors can have different shapes. However, in practice the method works well.

Optical generation of a circular harmonic filter for rotation and translation invariant optical pattern recognition

A new method to generate a circular harmonic filter with both rotation and translation invariance is described, which is based on a phase-shifted double-exposure (PSDE) technique. An expression for the peak correlation intensity at the origin of the correlation plane as a function of target orientation for the PSDE filter is derived. Experimental results confirming the theoretical predictions are provided.

Pattern recognition using reduced information content filters

Pattern recognition by optical spatial filtering procedures is discussed using general considerations with the objective of reducing the information content in the spatial filter. The achievement of this objective is very useful toward the wide application of spatial light modulators and also for facilitating distortion invariant recognition. The proposed novel approach is demonstrated by an example employing bipolar spatial filters for rotation invariant pattern recognition.

Properties of the circular harmonic expansion for rotation-invariant pattern recognition

Some general properties of the circular harmonic expansion relevant to their use for pattern recognition are derived. Both circular harmonic filters and Fourier-Mellin descriptors, which are used as the moments of circular harmonic functions, are considered. Expressions for the asymptotic energy in terms of the circular harmonic orders are derived and experimentally verified.

Rotation-invariant pattern recognition using the phase of circular harmonic filter correlations

At the center of the correlation between a circular harmonic matched filter and the object from which it was derived, there is a simple phase relation among the correlations from successive circular harmonic orders. This phase relation may be used to recognize the object of interest in a digital multiple matched filter method that is invariant under translation, rotation and changes of contrast of the targets. Experiments using photographs of space vehicles are presented.

SPACE-INVARIANT AND ROTATION-INVARIANT IN A GIVEN ANGLE RANGE PATTERN RECOGNITION AND THEIR FILTER DESIGN

Hsu et al have proposed a method of filter design for space-invariant and rotation-invariant digital pattern recognition in terms of circular harmonic expansion. Based on their method, we suggest a new filter design for space-invariant and rotation-invariant in a given angle range pattern recognition system. Although rotation-invariant angle is restricted to centain extents, the ratio of signal to noise might be increased, meanwhile we need not measure the phase factor of correlation function between target and recognition image in order to determine the rotation angle instead of directly measuring the absolute value of correlation function. We also give one simple example to demonstrate this design and compare it withthat suggested by Hsu et al.

A model for size- and rotation-invariant pattern processing in the visual system.

The mapping of retinal space onto the striate cortex of some mammals can be approximated by a log-polar function. It has been proposed that this mapping is of functional importance for scale- and rotation-invariant pattern recognition in the visual system. An exact log-polar transform converts centered scaling and rotation into translations. A subsequent translation-invariant transform, such as the absolute value of the Fourier transform, thus generates overall size- and rotation-invariance. In our model, the translation-invariance is realized via the R-transform. This transform can be executed by simple neural networks, and it does not require the complex computations of the Fourier transform, used in Mellin-transform size-invariance models. The logarithmic space distortion and differentiation in the first processing stage of the model is realized via "Mexican hat" filters whose diameter increases linearly with eccentricity, similar to the characteristics of the receptive fields of retinal ganglion cells. Except for some special cases, the model can explain object recognition independent of size, orientation and position. Some general problems of Mellin-type size-invariance models-that also apply to our model-are discussed.

Statistical performance of the circular harmonic filter for rotation-invariant pattern recognition

Circular harmonic filters of different orders are compared using a criterion of separability between the target and object output probability densities. A holographic computer-generated circular harmonic filter was used to recognize aircraft from an air photograph.