Computer Image Recognition Research Articles

시설멜론용 다기능 재배생력화 시스템;원격 로봇작업 시스템 개발

In this paper, a prototype tele-operative system with a mobile base was developed in order to automate cultivation of house melon. A man-machine interactive hybrid decision-making system via tele-operative task interface was proposed to overcome limitations of computer image recognition. Identifying house melon including position data from the field image was critical to automate cultivation. And it was not simple especially when melon is covered partly by leaves and stems. The developed system was composed of 5 major modules: (a) main remote monitoring and task control module, (b) wireless remote image acquisition and data transmission module, (c) three-wheel mobile base mounted with a 4 dof articulated type robot manipulator (d) exchangeable modular type end tools, and (e) melon storage module. The system was operated through the graphic user interface using touch screen monitor and wireless data communication among operator, computer, and machine. Once task was selected from the task control and monitoring module, the analog signal of the color image of the field was captured and transmitted to the host computer using R.F. module by wireless. A sequence of algorithms to identify location and size of a melon was performed based on the local image processing. Laboratory experiment showed the developed prototype system showed the practical feasibility of automating various cultivating tasks of house melon.

Sediment Fingerprints: A forensic technique using quartz sand grains

The analyses of 738 soil and sediment samples taken during the investigation of 20 forensic cases are combined in a database for retrospective comparisons. Quartz sand grain surface textures are combined into grain types which are analysed at one of five orders of classification. This classification also provides frequency data on the number of grain types found in a sample, the individuality of the quartz grain type assemblages and quartz grain type ratios. The technique, developed primarily as an exclusion mechanism, suggests possibilities, in some circumstances, that significant interpretations can be achieved at different orders of classification. The technique provides excellent exclusionary results and, when combined with other independent lines of analysis, is a powerful tool in forensic interpretation because it uses quartz as its discriminatory mineral. The technique also highlights the potential for computer image recognition analysis.

Locating Bird Roosts with Doppler Radar

Roosting birds of certain species can be agricultural pests, hazards to aircraft on takeoff and landing, and a purported health hazard. Locating roosts of pest bird species and estimating the numbers of birds using them is time-consuming field work, especially when use of different roosts changes seasonally. Since the middle of the previous century, radar has been used to observe early morning bird echoes, then called “ring angels”. Today, large Doppler radars designed for meteorological work can routinely observe bird roosts, even when birds fly at treetop height. Radar images can often locate all roosts within a certain distance from the radar and can provide an indication of the number of birds using each roost and the general location of their food sources. Single images from the lowest tilt angle (lowest elevation) of different radars show roosts in several areas of the country, and successive scans of a research radar across an area becomes an animated picture of the detailed spatial behavior of birds leaving the roost. Applying a computer image recognition technique, the Hough Transform, to single radar-derived images of bird roosts results in objective numerical estimates of roost location and other data. Data are shown comparing ground-truth visual counts of European starling and brown-headed cowbird departures with such quantitative radar-derived data. The radar correctly estimated the central tendency of flight speed of the birds (20 m/sec), time of morning flight (mean 13.2 min past civil sunrise), and roost location (modal error about 2 km). Sometimes (5.1% of identifications) the algorithm found a “roost” that could not be located by field observers; occasionally there were other sources of confusing echo such as vehicles or migrating birds.

Measuring coupled oscillations using an automated video analysis technique based on image recognition

The applications of the digital video image to the investigation of physical phenomena have increased enormously in recent years. The advances in computer technology and image recognition techniques allow the analysis of more complex problems. In this work, we study the movement of a damped coupled oscillation system. The motion is considered as a linear combination of two normal modes, i.e. the symmetric and antisymmetric modes. The image of the experiment is recorded with a video camera and analysed by means of software developed in our laboratory. The results show a very good agreement with the theory.

A simple method for measuring surface strains around cracks

A simple system has been developed to measure surface strains that occur during in situ deformation of mechanical test specimens. The system uses photolithographically deposited displacement markers and computer image recognition routines to determine in-plane displacements and strains from digital images. The strain calculating routines are integrated into a simple mouse-driven software package that facilitates the transformation from digital images to useful strain field information. Additional routines have been developed to determine crack tip stress fields and J integrals. Crack tip stress intensities have been calculated from strain maps obtained for traction-free cracks in stainless steel. The J integrals were found to be independent of contour and consistent with applied stress intensities. Crack tip stress intensities were calculated for bridged cracks in lamellar TiAl. The toughening effect of the bridging zones was determined by including the bridged region in the contours. Resistance curves generated from strain maps were consistent with those measured during mechanical testing.

眼球運動３成分の記録・解析

Examining eye movements is one of the most important neurotological tests. Thus, the accurate recording and analysis of eye movements is important in diagnosing vertigo and/or disequilibrium.There have been many methods of measuring eye movements. Such as direct examina-tion including cinecamera and VTR recordings, photoelectric examination, reflection methods including reflections from the cornea and from attachments to the eye, electronystagmography (electrooculography), electromagnetic recording, and methods using computer image recognition techniques.The computer image recognition method developed in our department is able to analyze three components of eye movements, horizontal, vertical, and torsional. The recording and analysis methods of this system are described and the practical application of this method is discussed.

Computer image processing and recognition

Computer image processing and recognition

Computer Image Processing and Recognition

A systematic introduction to the concepts and techniques of computer image processing and recognition is presented. Consideration is given to such topics as image formation and perception; computer representation of images; image enhancement and restoration; reconstruction from projections; digital television, encoding, and data compression; scene understanding; scene matching and recognition; and processing techniques for linear systems.