Abstract: A pattern checking device receives scanning data obtained by raster scanning objects within predefined window regions with a photoelectric converter, converts the scanned data into pixel binary data, and evaluates the binary data in accordance with predetermined criteria to produce an output signal indicative of the evaluation. The evaluation is performed in at least two steps. In the first step, a primary decision circuit compares individual data from each window region in a group to upper and lower thresholds, and outputs primary decision result signals for each window region. In the second step, a secondary decision circuit compares all the primary decision result signals for all the window regions in the group with previously stored decision data for the group, and outputs a pattern check signal indicating whether the pattern appearing in the group of window regions constitutes a good pattern.

Abstract: A signature verification system includes means for obtaining acceleration data and pressure data related to a given signature and means for comparing that data with reference data previously obtained from an identified signer. The correlations between segments of a reference pressure signal and a sample pressure signal are computed. This process involves successively measuring the correlation as the two segments are shifted with respect to each other. The amount of the shifts that produce the maximum correlations between the pressure signal segments is stored in memory, and these shift values are utilized to determine the correlations of respective pairs of segments of the acceleration data. Since the acceleration data is not independently shifted in computing the correlations, substantial processing time is saved. In addition, the overall performance of the signature verification system is improved both in terms of processing time and in error rate as compared to the prior system.

Abstract: A pattern discriminator receives scanning data obtained by raster scanning objects within predefined window regions with a photoelectric converter, converts the scanned data into pixel binary data by using a threshold value dependent upon the scanning position and window region and evaluates the binary data in accordance with predetermined criteria to produce an output signal indicative of the evalution. Data specifying the shapes of the desired window regions approximated by polygons is obtained by specifying the apexes of the polygons and a programmed CPU computes the coordinates of each picture element or pixel on the lines connecting adjacent apexes along each window region perimeter. A memory stores the apex and coordinate data which defines the starting and ending points of each window region. Polygonal masks within each window and sub-mask polygonal window regions within each mask can be defined in a similar manner.

Abstract: In an improved method for inspecting crimp bonded terminals, instead of inspecting visually, contour images of crimp bonded terminals are picked up and scanned to provide necessary data and processing operations are performed on the data to make a decision automatically on the acceptance or rejection of the crimp bonded terminals, thereby elevating the inspection efficiency and improving the accuracy of deciding the acceptance or rejection of the crimp bonded terminals.

Abstract: A method of diagnosing a combustion state in a furnace of a boiler wherein the shape of the root part of a flame (namely, the part closer to the fuel jet port of a burner) is picked up by an image sensor or the like, and the combustion state is diagnosed from the shape. Concretely, the relationships between the shapes of the root parts of flames and the combustion states of the furnace, e.g., the amount of CO or the amount of NOx are experimentally or empirically ensured and are stored in advance, one of the patterns of the stored flame shapes is selected on the basis of the shape of the root part of the flame actually observed, and the combustion state of the furnace is judged the combustion state of the selected flame shape.

Abstract: A signature verification method is based on a comparison of the dynamics of a reference and a sample signature. Acceleration and pressure signals produced by a known person when writing his or her signature are stored and used as a reference signals. Then, at a later time, a person whose signature is to be verified writes his or her signature to produce acceleration and pressure signals that are compared to the reference signals. The process of comparison involves segmenting the two sets of signals to facilitate identifying regions of high probable correlation and then correlating corresponding segment pairs. Segmentation is based on pen lifts which represent reproducible timing marks in the signatures. According to the disclosed method, a pen or other writing instrument is used which produces a signal representative of the first time derivative of the pressure forces exerted on the stylus of the pen. The second time derivative of the pressure forces is computed from the measured signal.

Abstract: A signature verification system includes means for obtaining acceleration data and pressure data related to a given signature and means for comparing that data with reference data previously obtained from an identified signer. The correlations between segments of a reference pressure signal and a sample pressure signal are computed. This process involves successively measuring the correlation as the two segments are shifted with respect to each other. The amount of the shifts that produce the maximum correlations between the pressure signal segments is stored in memory. The corresponding segments of acceleration data are shifted with respect to one another by an amount which is .+-.1 time units of the stored shift values during subsequent correlations of the acceleration data.

Abstract: A pattern discriminator receives scanning data obtained by raster scanning objects within predefined window regions with a photoelectric converter, converts the scanned data into pixel binary data by using a threshold value dependent upon the scanning position and window region and evaluates the binary data in accordance with predetermined criteria to produce an output signal indicative of the evaluation. A memory stores data which represents the window regions in a compressed arrangement indicating the ordinate and abscissa coordinates of the beginning and ending points or pixels of each window region location on each scanning line, along with corresponding information representing the particular threshold value to be used for each window region segment on each scanning line.

Abstract: In a pellet pattern recognition method, a pellet pattern recognition apparatus is used which includes a table for determining a pellet pick-up priority order, pellet memory table for storing data of good and bad pellets as well as data representing an "already picked-up" state and a "not picked-up" state, and position memory table for storing a positional shift of each of the good pellets within a pellet pattern recognition range away from the picked-up pellet. When the good pellets are to be sequentially picked up, the shape identification and position detection of the pellets within the pellet pattern recognition range are effected by calculating the correlation coefficients. In this case, calculation is performed with respect to the pellet array on the basis of the results of the shape identification and position detection and initially set reference data, and the good pellet is stored, as the corresponding coordinate data, in the pellet memory table.

Abstract: A neighborhood transformation stage is designed to perform filtering operations on an image matrix. A group of pixel values in the image matrix are accessed. A series of comparators are used to compare the values of neighboring pixels with the value of a given pixel in the group. The outputs of the comparators are combined to form a vector to address a programmable look-up table. The addressed output of the look-up table is used as a control signal for a selector which either maintains the value of the given pixel or transforms it into another value associated with a neighboring pixel whose value has been ranked in a predetermined order by a rank sorter device used to compare the values of the neighboring pixels with each other.

Abstract: A process for thinning an image, comprising the steps of defining an image in terms of a plurality of discrete picture elements (pixels), convolving each horizontal string of pixels of the image with itself shifted a predetermined number of pixels, weighting each shifted position of each horizontal string, selecting certain of the pixels of the weighted horizontally convolved pixel strings which correspond to predetermined descriptors, convolving each vertical string of pixels of the image with itself shifted a predetermined number of pixels, weighting each shifted position of each vertical string, selecting certain of the pexels of the weighted vertically convolved pixel strings which correspond to predetermined descriptors, adding thee selected pixels from the weighted vertical and horizontal convolutions and the pixels of the original image, retaining those pixels which are common to the original image and the selected pixels of both the horizontal and vertical weighted convolutions, discarding those pixel

Abstract: A gray scale filter for image data is described which forces a least significant bit of a current pel equal to a most significant bit of the current pel when the most significant bit of the current pel is equal to the most significant bit of a previous and a next pel on a scan line of an image. The invention is embodied in a logic structure employing exclusive ORs to make comparisons between the previous, next and current pels and logic gates to gate the value of the most significant bit of the most current pel to an output representing a filtered least significant bit of the current pel. It should be noted that the value of the most significant bit of the current pel is not changed since only changes between gray-white and white or gray-black and black are filtered by the present invention.

Abstract: A method for recognizing complex handwritten character set employs spatial position recognition in which characters of a first set are registered in a first region and characters of a second set are registered in a second region. The second region may be included within the first region. The invention is particularly useful for distinguishing Kanji from non-Kanji character sets in a data input system.

Abstract: A pattern recognition system which detects line bifurcations and line endings, denoted minutiae, in a pattern of lines such as are found in a fingerprint is disclosed. In one embodiment, an image of a pattern is focused on a conventional matrix of image sensors or pixels. An interface circuit serially reads the voltages from the pixels and applies them to an automatic, programmable threshold detector which optimally selects a cutoff voltage and outputs a binary signal representative of the presence of a point to a minutiae detection circuit. One part of this circuit electrically reforms the matrix on a first-in-first-out basis and simultaneously forms a smaller submatrix or window. Another part of the detection circuit determines firstly if a point is present in the central array of the window, and if so, secondly whether no more than two points, which are also contiguous, are present in the peripheral part of the submatrix.

Abstract: A method applicable to a character recognition system is disclosed which assigns direction codes to a number of boundary picture elements contained in a two-level character pattern. The direction of connectivity of a boundary picture element observed is fractionized to minimize the error due to the quantization of the direction. The direction codes are converted into those which correspond to connectivity directions which should be finally grasped. Such direction codes allow strokes to be extracted quite faithfully to the original character pattern.

Abstract: A computer algorithm and apparatus for automated image input (including recognition), storage and output (including image generation) is described. Each image is transformed into a unique binary number and then stored as such. This unique binary number is obtained by adding unique binary value of each pixell present in that image. Means for processing handwriting and colored images are also disclosed. Image recognition and matching takes place by comparing the binary value of the new image received against all images stored in the descending order of difference in binary values. Thus the computer is able to recognize bad handwriting even when the difference between the ideal or stored samples on the one hand and the new image is substantial without consistency. The computer also stores data about its errors as well as corrections received from the user. For this and other reasons each user has a unique number.