RU2173880C2 - Method and device for separating attributes to identify images of objects - Google Patents

Abstract

FIELD: identification of images. SUBSTANCE: method boils down to conversion of optical signals into electric video signals, sampling of video signals, separation of object image into discrete elements, image scanning, separation and counting of discrete elements. All image discrete elements are scanned by electronic window in the form of background frame accommodating in its center object image viewport; dimensions of the latter are comparable with those of object image; in separation and counting procedures discrete elements of object image viewpoint with electric signal values other than those found among discrete elements of background frame are assumed as discrete elements of object image. Device implementing proposed method has position signal shapers 1 and 2, switches 3, 4, and 6, matrix- type photoelectric converter 5, video signal sampling circuit 7, memory 8, timing device 9, AND circuit 10, and counter 11. EFFECT: enhanced reliability of polychromatic object image separation without pre-filtering. 2 cl, 5 dwg

Description

 The present invention relates to metrology, automation and technical cybernetics and can be used in the development of automatic analysis and classification of images.

 A known method of generating features for recognizing images of homogeneous structures, which consists in line-by-line scanning of images of objects or structures, generating signals of the intersection of the object contour with a scanning beam and subsequent analysis of the resulting video pulses corresponding to the chords of the scanned objects / 1 /.

 This method has low recognition reliability, because can be successfully used only for recognition and analysis of structures that are formed by elements with clearly defined boundaries, the formation of signs of grayscale images in this way is impossible.

 This method can be implemented using an image recognition device containing a first memory block connected via a switching unit connected to a shift register to a first counter unit, a second memory block connected through a shift register to a control unit, a classification unit for recognizing pattern spectra, connected with the first counter unit and the control unit, the second counter unit connected to the first memory unit and to the control unit, and the third counter unit connected to the second memory unit and control Nia / 2 /.

 The disadvantages of this device are its structural complexity, the ability to work only with black and white images.

 Closest to the proposed technical essence is a method of identifying features for recognition of objects, which consists in dividing the projection of the image of objects into discrete elements by repeatedly projecting at least two rasters, discrete deformation of raster images, and selecting and counting as images of image areas of intersection of rasters / 3 /.

 However, using this method, it is impossible to select images of objects with alternating contrast, which is most often found in natural structures.

 This method can be implemented using a device for preliminary processing of the image to be recognized containing a photoelectric converter for scanning the original image, generating signals at the second output and time-sequential electrical video signals at the first output, to which are connected in series a circuit for sampling video signals, a logic circuit for the first entrance and a set of counters; a circuit input for generating positional signals determining the position of each binary video signal on the image plane and a memory input using positional signals as addressable, the output of which is connected to the second input of the logic circuit and the circuit output for generating positional signals is connected to the second output of the photoelectric converter with one of the inputs of the set of counters / 4 /.

 The disadvantage of this device is the limited number of implemented operations for image processing.

 The objective of the invention is to increase the reliability of highlighting a multi-color image of an object with alternating contrast without first improving image quality (filtering signals) in a complex background by highlighting certain dimensions in the field of view of an object whose color (grayscale) gamut of the image differs from color (grayscale) gamma ambient background.

The problem is achieved in that in the present invention in a method consisting in converting optical signals into electrical video signals, sampling video signals, dividing the image of objects into discrete elements, scanning an image, extracting and counting discrete elements, scanning across all discrete image elements is performed by an electronic window in in the form of a background frame, inside of which in the center there is an electronic image window of an object (OIO), and the sizes of the OIO are commensurate with the dimensions and siderations object, by isolating and counting as discrete object image elements received OAM discrete elements with such values of electric signals among the signals which are not discrete elements background frame;
into a device for preliminary processing of the image to be recognized, comprising a photoelectric converter, a positioner, a video sampling circuit, a memory, an AND circuit, a counter, the memory output being connected to a first input of an AND circuit, the output of which is connected to a second input counter, additionally introduced a temporary device, the first switch, the second driver of positional signals, the second switch, the third switch, and the output of the second driver signal is connected to the second input of the second switch, the first input of which is connected to the output of the first positioner, the third input to the first output of the temporary device and the input of the first switch, and the output is connected to the second input of the third switch, the first input of which is connected to the output of the photoelectric converter matrix type, the output of the third switch is connected to the input of the video sampling circuit, the output of which is connected to the second memory input, the first input of which is connected output of the first switch, the third input - to the second output timing device, the fourth input - a third output timing device and the first input of the counter, the fourth output timing device coupled to the second input of the logic "AND", the counter output is the output device.

 The positive effect is ensured by isolating and counting discrete elements of the matrix section of the photoelectric converter, commensurate with the dimensions of the desired object, with the values of electrical signals that are not among the signals of the discrete elements of the background frame surrounding the image of the object, and preliminary improvement of image quality is not required (filtering of the photoelectric signals converter), because as a result of the operations performed, the image is excluded (deleted) from the plot on with any noise.

 The proposed method is as follows.

 The matrix-type photoelectric converter receives optical signals from the field of view containing the desired object, converts optical signals into electrical video signals and splits the plot image into discrete elements. The video signals are sampled and the signals from the photoelectric converter matrix elements are sequentially viewed by an electronic window in the form of a background frame, inside of which in the center there is an electronic image window of an object, and the dimensions of the image window of the object are comparable with the size of the image of the object. From discrete elements of the image window of an object, elements with signal values that are present among signals from discrete elements of the background frame are excluded. This operation selects an image of an object having alternating contrast with respect to the surrounding background and sizes corresponding to the dimensions of the desired object. As a sign of the image of the object, for example, the area, discrete elements of the object section remaining after exclusion are counted.

 This technical solution is illustrated by graphic materials.

 In FIG. 1 shows the location of the areas of the electronic window on a fragment of the plot image formed by the photoelectric converter, where 1 is the background section (background frame), 2 is the object image window (OIO). On this fragment containing images of the desired object and background, for recognition, it is necessary to select the image of the object. The signal values of the background frame are determined. The background frame contains signals with amplitudes conventionally shown in FIG. 2. Eliminate (erase) from the OIO image the matrix elements of the photoelectric converter with the given signal values. As a result, the image of the desired object remains in the OIO (Fig. 3).

 In FIG. 4 shows a block diagram of a device with which it is possible to implement the proposed method, FIG. 5 is a sequence diagram of the operation of a temporary device.

 The proposed device operates as follows. When power is applied, the signal from the first output of the temporary device enters the input 3 of switch 2 (4) and sets it to the position when the OIO addresses defined in the second positioner are received at the second input of switch 3 (6) through the second input and output of switch 2 OIO (2) in accordance with the law of scanning, and the switch 1 (3) is set to state "1", in which signal "1" is received from its output to the first memory input.

 At the second output of block 9, a memory recording signal is generated, which is fed to the third memory input (8), which allows recording information. At the third output of the temporary device (9), a pulse arises that sets the device to its initial state: reset (write "0" to all elements) at the fourth memory input and counter at input 1. The signal "0" from the fourth output of block 9 is fed to the second input logic circuit (10), which is an AND circuit, and blocks the operation of the counter (11).

 The analyzed image is projected onto the matrix of the photoelectric converter. At the addresses coming from block 2 through switch 2 to the second input of switch 3, block 6 connects the photoelectric converter matrix elements belonging to the OIO to the video sampling circuit (7), in which the analog signal from the photoelectric converter is converted to a digital code. In block 8, there is a recording of level "1", acting at the input 1 from the output of the switch 1, at the addresses received in the memory from the output of the video sampling circuit. As a result of the interrogation of the OIO elements in the memory at addresses equal to the amplitudes of the signals from the matrix elements of the photoelectric converter corresponding to the OIO, "1" is written, that is, the values of the amplitudes of the OIO signals are determined through the values of the memory addresses.

 At the end of the interrogation of the elements of the OIO, at the first output of the temporary device, an impulse of interrogation of elements of the background frame appears, which goes to the third input of block 4 and switches the state of switch 2 to the mode of connecting the first driver of position signals of the background frame (1) to the second input of switch 3, and switch 1 to state "0".

 Now, at the addresses arriving at the second input of block 6 from block 1, the switch 3 connects to the video sampling circuit the photoelectric converter matrix elements belonging to the background frame. After the sampling scheme of the video signals, the values of the address of the background frame appear at the second memory input, at which the level "0" acting on the first input is written to the memory. In this case, the state of only those memory cells changes in which “1” was recorded in the previous step (polling of the OIO elements), i.e. information is erased from the memory at addresses expressed by signal amplitudes that are the same for the OIO and the background frame.

 At the end of the interrogation of the elements of the background frame, “1” is written to the memory only at those addresses whose values are equal to the amplitudes of the signals from the elements of the OIO and are not among the amplitudes from the elements of the background frame.

 At the same time, a read signal from the second output of the temporary device appears at the third memory input, a signal "1" is generated at the fourth output of the temporary device, which is fed to the second input of block 10 and allows the counter to count single pulses arriving at its second input. The pulse of the first output of the temporary device switches the switch 2 to its original state when the OIO addresses arrive at the second input of the switch 3. If a pulse arises at the second memory input with an amplitude equal to the address of the memory cell in which "1" is stored, a positive pulse is output from the memory output to the first input of the "And" logic circuit, and from the output of block 10 to the second input of block 11, where counting unit pulses.

 The counter can be performed, for example, on the K155IE2 chip, the logic circuit “I” is K155LA1, the switches are K176KT1, the memory is K155RU5, the sampling circuit is an analog-to-digital converter K1107 PV1, the photoelectric converter can be performed, for example, on a CCD / 5 /, the temporary device is like a timer 555/6 /, and the positioner of the signal signals is similar to the prototype / 4 /.

 Thus, the use of the proposed method and device allows you to select the image of the object with dimensions corresponding to the dimensions of the desired object, having a contrast with respect to the surrounding background.

 The proposed method and device increase the reliability of object recognition, provide without filtering signals the ability to recognize grayscale objects with alternating contrast against a complex background in the presence of interference.

 The inventive method of identifying features for recognizing images of objects and a device for its implementation are tested with great efficiency in the Hermes complex.

Sources of information
1. A.S. USSR N 748454, class. G 06 K 9/00, 1980 - analogue.

 2. A.S. USSR N 650087, class G 06 K 9/00, 1970 - analogue.

 3. A.S. USSR N 987643, class G 06 K 9/30, 7.01.83 - prototype.

 4. A. z. Japan N 64-6508, cl. G 06 F 15/64, 15/66, G 06 K 9/20, 9/36 - prototype.

 5. L.I. Khromov, N.V. Lebedev, A.K. Tsytsulin, A.N. Kulikov. Solid state television. M .: Radio and communications, 1986.

 6. M. Tuli. Reference manual on digital electronics / M.: Energoizdat, 1990, pp. 36-40.

Claims (2)

 1. A method of extracting features for image recognition of objects, which consists in converting optical signals into electrical video signals, sampling video signals, dividing the image of objects into discrete elements, scanning the image, extracting and counting discrete elements, characterized in that the scanning of all discrete image elements is performed electronically a window in the form of a background frame, inside of which in the center there is an electronic image window of the object, the dimensions of the image window being CTA commensurate with the size of the object image, by isolating and counting as discrete object image elements taken discrete elements of an image window with such values of electric signals among the signals which are not discrete elements background frame.  2. A device for extracting features for image recognition of objects, containing a photoelectric converter, a positioner of the signal, a sampling circuit for video signals, a memory, a logic circuit And, a counter, the memory output connected to the first input of the logic circuit And, the output of which is connected to the second input of the counter, characterized in that a temporary device, a first switch, a second positioner, a second switch, a third switch are inserted into it, the output of the second driver I positional signals are connected to the second input of the second switch, the first input of which is connected to the output of the first positioner, the third input to the first output of the temporary device and the input of the first switch, and the output is connected to the second input of the third switch, the first input of which is connected to the photoelectric output matrix type converter, the output of the third switch is connected to the input of the video sampling circuit, the output of which is connected to the second memory input, the first input of which is connected inen with the output of the first switch, the third input with the second output of the temporary device, the fourth input with the third output of the temporary device and the first input of the counter, the fourth output of the temporary device is connected to the second input of the logic circuit And, the output of the counter is the output of the device.

Images