Method of detecting objects

FIELD: television systems.

SUBSTANCE: method comprises subtracting reference and current images, breaking the image series to be processed into fragments, and converting the characteristic features of the images into signals. The signals from one of the images are recorded as reference ones and are compared, e.g., by subtracting, with corresponding current signals, and, after the threshold processing, the difference signals obtained are converted into the binary signals for control of spatial filtration . As a result, the fragments of the current image, for which the control signals exceed the threshold, are transmitted, whereas the fragments, for which the signals are equal or less than the threshold value, are suppressed.

EFFECT: enhanced quality of the object image.

11 cl, 14 dwg

The invention relates to image processing and can be used for automatic detection, selection of images and measurements of parameters of objects in complex backgrounds. The proposed method for the detection of objects can also be used for monitoring of protected areas and the premises against unauthorized entry with registration, if necessary, the image of the object of the offender; in automatic recognition and tracking, and other image processing systems.

Analogue of the invention according to the functional purpose - finding and selecting objects - a method based on the effect of straightening the current structures of the type metal - dielectric-semiconductor - liquid crystal (MDPC) [1, str-236].

When the power of this bipolar structure (e.g., sinusoidal) voltage is straightening passing through the structure of the current, which causes a DC component of the voltage across the liquid crystal layer (LC). The measurement of the photo-EMF that occurs when light PWMs with one or the other side plate photopolarimetric, indicates the presence of curvature of the energy bands of a semiconductor on its borders. The field curvature playing electric regard, the role of diodes connected towards each other. When lighting one of the diodes, so what. the corresponding side of the structure, the reverse resistance of the diode decreases, which leads to the asymmetry of the structure and to the emergence of a rectifying effect.

The rectifying properties of the structure determine the optical response. After application of the AC voltage is rectified component in the first moment is distributed inversely proportional to the capacitances of the dielectric layers With_dLCD With_LCD. With regard to their real values it is clear that at this point, most of the voltage drop in the layer LC, and only after a time equal to the time of discharge From_LCDthrough their own resistance R_LCDit will take its stationary value, which is very little. Thus, when changing the AC voltage is the roll component of the rectified voltage on the LCD layer. This change occurs when the lighting patterns, decreases sharply when the resistance of the illuminated diode and the resistance of the volume of semiconductor crystal R_PP.

The duration of the surge voltage is determined by the discharging time of the LCD layer and is usually 1... 10 MS for used LCD (resistivity 10¹⁰Ohm· cm). The magnitude of the supply voltage can always be chosen so that the LCD was on the verge of transition; then the surge, the extended structure in the PTO is t her lighting, will determine the nature of the optical response. Thus, the optical response exists only on the change of illumination and therefore patterns have specified dynamic properties.

The sensitivity of these structures to non-stationary (moving, flickering) objects by 2-3 orders of magnitude higher than for stationary (background)that detects and discriminates the image non-stationary objects in complex stationary backgrounds, as the device only responds to moving and flickering (changing brightness) of the object.

The disadvantage of this method is the ability to detect only moving (spatial nonstationarity) or changing the brightness, “flickering” of objects (temporal nonstationarity). Hence another major drawback - stringent requirements for spatial and temporal stabilization of the background, which is especially difficult when working on a natural, “natural” background.

The closest analogue prototype is a method for detection and selection of images from complex backgrounds, based on the subtraction images [1, str-245], [2, str-112].

In accordance with this method write (and, if necessary, overwrite) into memory and is defined as supporting one of the current images. When the subtraction of the reference and subsequent current image at the output of which is jut a differential image. If successful the target environment (FCO) changes have occurred, such as the appearance of a new object, the resulting difference image corresponds to the image of this object. Constant component (background) is excluded.

This method allows you to find almost any object on complex backgrounds, but at the same time, it has several significant disadvantages, including:

- low quality of the selected image object, as when subtracting each point of the image of the object is formed as the difference between “background object” and therefore, there is a strong distortion - “the breaks”, “fade”, “spotting”, and others;

- high stability requirements of the background and the relative shifts “system-background”, because when you change the background and relative shifts a mismatch occurs between the signals in the reference and current frames and subtracting the difference signal of the background vanishes, that is, noises appear.

The task of the invention is to develop a method that ensures high quality of the obtained image of the detected object.

The invention consists in that the known method of detection of objects containing the preprocessing of the signals are represented by a temporal sequence of images of a scene, which may cause the object is, memorizing a reference describing the scene image signals and the subtraction from the corresponding current of the respective reference signals, after preprocessing, each presented in optical or videosignals form, for example in the form of a television signal, the image is considered a sequence of images divided into fragments, and then measure the magnitude of the characteristic values of each fragment, for example the total optical density of the fragments of the optical image or the total of the amplitudes of the video fragments for the television image, and remember the corresponding signals, the corresponding values of the signals corresponding to fragments of the current image, subtract the values of the corresponding reference signals of the respective fragments and the resulting differential signal is compared with a specified threshold the value of the signal, and then generate the control signals, by means of which provide granular filtering time sequence of the current image in one of the fragments, the absolute values of the differential signals which exceed the threshold value, generate control signals, transmits the corresponding image signals, and for fragments, the absolute values of the differential signals are less than or ranipauwa value form the control signals, blocking the corresponding image signals.

In addition, parts of the image choose the same configuration and of equal area with the dimension k of the pixels on each fragment, where K_i=1,... , n, n≤ N, and N is the number of pixels in the image, and the image select different configuration space of dimension k_ipixels on each i-th fragment, where k=1,... , n, n_i≤N/2, and N is the number of pixels in the image, and the image chosen partly the same and partly different in configuration space with dimension k_ipixels on each i-th fragment and each fragment k_i=k.

In addition, for each image of the time sequence of images chosen by the same corresponding parts of the images, and also fragments of images to choose different size and configuration, and size and configuration of the fragments chosen, taking into account the possibilities when comparing the respective current and reference image hosting and matches the configuration of the set of corresponding fragments smaller area on the corresponding piece of the big square. In addition, preprocessing is carried out in the form of spectral filtering of the image.

As a researcher who as the reference select signals characterizing magnitudes of signals of the image fragments, or, for example, each previous to subsequent, or, for example, the first image for all temporal sequence of images. In addition, as signals for each image slice select signals corresponding to the value obtained by the average processing signals obtained by measuring the characteristic values of the corresponding fragments for a given number of images considered a temporal sequence of images, as well as signals corresponding inverted images of the respective signals. The parameters of the scene and the object is obtained by measuring the number, order and the brightness of the fragments placed on the object and specified areas of the scene. This method in comparison with the prototype provides:

- High quality image, the selected target object corresponding to that which provides the perceptual system. Moreover, the filtering may be performed on the optical carrier, i.e. directly after the input lens with maximum resolution, color retention, etc.;

- Reduced sensitivity to instability of the background and system;

- Possibility of measuring the geometric parameters of the object and the background environment by counting the number of fragments in the image of the object and the background measurement is of ontrast background object, etc.;

- Possibility of choosing the optimal balance between detecting ability, the ratio signal/noise ratio and the accuracy of determination of existence and boundaries of the object by selecting the dimensions of the image fragments and others;

- The ability to pre-recognition and classification of images of the object by measuring a similarity measure characteristics of the fragments with typical (generic) image belonging to a given class of interest and recorded as a reference.

In figure 1(a-d) shows the results of processing by the subtraction method (prototype). Here in the figure 1(a, b) presents the reference and current frames, respectively; figure 1(C) shows the selected object in stable conditions; figure 1(d) shows the effect of relative shifts “system-background”; in figure 1(d) presents the results of the selection object using the operations of the method. It should be noted (figne shown)that when you filter directly after the lens is stored and color images.

In figure 2(a-d) shows an example of processing of the proposed method (slice boundaries shown for clarity). The numbering of the fragments (5× 5) sequential, left-to-right, top-to-bottom.

Here: figure 2(a, b) shows the reference and current frames, respectively; figure 2(b) provides a graphic representation of the different is subjected to brightness fragments (differential signals); in figure 2(d) shows the selected object. The accuracy of the boundaries is determined by the size of the fragments.

In figure 3(a-d) as an example of the implementation of the proposed method shows a variant of the circuit device in parallel with an optical input, fragmentation and image output.

Here in the figure 3(a) shows a diagram of the device with parallel optical fragmentation and output the selected image object directly after the input lens. In figures 3(6) presents options for units of the scheme.

The figure 4 shows an example implementation of the method using a processing circuit of the video signal for the television image.

The input lens 1₁(3) optically connected with the optical divider 2, made in the form of beam-splitting prism spectral optical filter 3 and the block containing the fiber optic focon 4 [2, p.114-115, Appendix 1], optically associated with the photosensitive matrix 5, made in the form of a CCD type FSS MB with frame-to-frame subtraction [2, str-112, Appendix 2], is connected electrically with the block 6 threshold processing executed in the form of the amplitude selector on the basis of, for example, chip SA [3, str]. Unit 6 thresholding is electrically connected to the unit 7 the generation of control signals, made for example in the form of chips CTS G or 133 series [3 : 272-274] and ele is traceski connected to the unit 8, the spatial filtering of fragments, which used electrically driven spatio-temporal light modulator liquid crystal (PVSEC) [4, p. 1, 11. Appendix 3]. The optical unit 8 is connected to the input and output lenses 1₁and 1₂and an optical splitter 2, and is electrically connected also to the power output matrix 5.

In figure 3(b) shows the options for building a site that is made to figure 3(a) in the form of a focon 4 and the plates 5 and provides smooth adjustment of the sizes of the fragments, which “breaks” the image. In figure 3(6) between the focon 4 and the sensor 5 is entered optically associated electro-optical Converter (EOC) 9, providing a smooth zoom image [5, p. 1. Appendix 4]. In figure 3(C) image intensifier 9 through the optical lens 1 is associated with the matrix 5 [6, p. 1, 7. Appendix 5].

The device operates as follows. The input lens 1_iforms the image coming from the image source in the form of a real object, from the Tube, or cathode ray tube (CRT) (primary data - video) in the input plane of the focon 4 and PWM LCD 8. After the large-scale conversion using focon 4 image is fed to the input matrix 5 FSS MB. Matrix 5 performs the operation of frame-to-frame subtraction (MB), i.e. registers the signal characteristics from the images, representing, in this case, the integral (total) intensity values of the fragments, writes the reference signals and through interframe interval subtracts the reference and the corresponding current signals of image fragments, forming the output differential signals. Differential signals after the processing in block 6, the threshold processing unit 7 the generation of control signals is converted into a binary (0,1) control signals which are fed to the electrical input electrically-controlled LCD modulator 8 mode optical shutter.

The elements of the LCD modulator 8, which receives signals equal to 1, open them, reflect (skip for LCD transmissive) of an image formed by the lens 1 on the LCD modulator.

The output of the circuit using the lens 1₂form the image onselectionchange from a background object. Note: when using a transmissive LCD modulator 8 output lens is removed from the schema.

The dimension of the portion of the image, i.e. the number of resolution elements in the fragment, the intensity of which is summed up one photosensitive element CCD, determined in this case from (N₁/N₂)m, m≥ 1, where N₁- the number of resolution elements in the input image (frame), N₂the number of sensitive elements in AP is rture CCD, a m is the scale factor of the transformation.

Ways to build the schema shown in figure 3(b)are used for changing the size of the fragments and harmonize formats of the input image and the aperture matrix 5.

In figure 3(d) shows a diagram similar to the diagram in figure 3(a), but in which the block fragmentation 4 exclude, that is, the selection in this case is equal to one element resolution (pixel). Thus the circuit operates in the normal mode frame-to-frame subtraction (prototype) using the matrix 5, made on the basis PSSPS [2, str-112, Appendix 2], but the inclusion of blocks 2, 6, 7, 8 leads to an improvement of the image quality is no “spot”, a higher resolution and color retention (since the image of the object is allocated immediately after the lens 1, the resolution of which is substantially higher resolution FSS MB matrix. The scheme works in a similar way shown in figure 3(a), but without surgery fragmentation performed by the block 4. Distortion introduced LCD modulator 8 [4, 1, 11, Appendix 3], insignificant, given its high optical quality.

Comparative results are shown in figure 1(C, d). Figure 1(b) is the image under normal interframe difference (prototype); figure 1(d) - using the schema in figure 3(d).

The main advantage is that the image quality, sharpness of boundaries due to the minimum size of f is amenta.

The main limitation for the application - high demands on spatial stability of the system and the background, as in the prototype.

In the figure 4 as another example shows a case where the device map for the implementation of the proposed method by processing the video signal, where the fragments of the broken line of the television image. Such a device can be used for detection in complex backgrounds high-speed point-like objects (for example, the task of multitracking) and allows to eliminate in comparison with the prototype breaks trajectories arising from the subtraction images of the background object.

The schema contains the generator horizontal 10, performed on the chip KGF and electrically connected to the differentiating circuit 11, which is electrically connected to one-shot 12, which is made on the chip CIE and the blocking oscillator 13, performed according to the standard scheme [8, str-214], which in turn is connected to one-shot 12.

Follow electrically connected to the blocking oscillator 13 of the electronic key 14, which is made on the chip KRKI and further, electrically connected blocks: the integrating circuit 15, the delay line 16 (CRHC) and the inverter 17 (D), the adder 18, which is made on the chip KIM, the amplitude selector 6, performed on the chip CSA [3, str], and therefore the and control 19 on the chip CIE and television cathode ray tube CRT 20. Amplifier 21 on the chip 171982 electrically connected with the electronic key 14 and the CRT 20 through the delay line 22 (CRHC).

The elements of this scheme are given in [7].

The scheme works as follows.

With generator horizontal 10 pulses beginning of the sweep line is coming to a differentiating circuit 11, forming pulses of positive and negative polarity. The positive pulse triggers the one-shot 12 connecting the supply voltage to the blocking generator 13 at the time of passage of the line through the key 14. Negative pulse triggers the blocking oscillator 13. The pulses of the blocking oscillator periodically open the key 14, separating the fragments of the video signal coming from the amplifier 21. Quantized video signal is supplied to the integrating circuit 15. The resulting signals are received from the integrating circuit 15 to the adder 18 through the delay line 16 and an inverter 17. From the adder 18, the pulses arrive at the limiter 6, and from it to the control unit 19. The delayed signal from delay line 22 is supplied to cathode ray tube 20. The pulses from the control unit 19 open cathode-ray tube, resulting in a displayed image of the part (fragment) of the video signal, which carries information about the object.

LITERATURE

[1] Ali, Dossent and other “Prostranstvo the s light modulators”,Moscow, ″ Radio and telecommunications″ , 1987, 320 S.

[2] Waseco, Vigilon “Microcircuitry BIS on charge-coupled devices”, Moscow, ″ Radio and telecommunications″ , 1988, 161 S.

[3] “the Reference book of the ham-constructor, Moscow, Radio and communications, 1990, 623 S.

[4] Terroelektric liquid crystal over silicon spatial light modulators - principles, practice and prospects. “Optical Society of America, Washington DC, 1997 g. pp. 76-88.

[5] Applied Optics, 1986, T. 25, No. 14, 2306-2310 C.

[6] Wppages, AAC, Astrachem “Electro-optical converters. Status and trends”. International centre for microelectronics and night vision “Orex”, Moscow; state unitary enterprise NPO “Orion”, Moscow.

[7] Wphsc “Reference item and analogues domestic circuits”, Novosibirsk, 7, Aug. 2000

[8] “the encyclopedia of physics”, Moscow, “Soviet encyclopedia, 1988, 704 S.

1. The method of detecting objects, comprising preprocessing the signals are represented by a temporal sequence of images of a scene, where the appearance of the object, storing a reference describing the scene image signals and the subtraction from the corresponding current of the respective reference signals, characterized in that after preprocessing, each presented in optical or videosignals form, for example in the form of a television signal, the image under consideration of the sequence of the images is divided into fragments, then measure the magnitude of the characteristic values of each fragment, for example the total optical density of the fragments of the optical image or the total of the amplitudes of the video fragments for the television image, and remember the corresponding signals, the corresponding values of the signals corresponding to fragments of the current image, subtract the values of the corresponding reference signals of the respective fragments and the resulting differential signal is compared with the specified threshold value signal, and then generate the control signals, by means of which provide granular filtering time sequence of the current image in one of the fragments, the absolute values of the differential signals which exceed the threshold value, generate control signals, transmits the corresponding image signals and for fragments, the absolute values of the differential signals are less than or equal to the threshold value, generate control signals, blocking the corresponding image signals.

2. The method according to claim 1, characterized in that the portions of the image choose the same configuration and of equal area with the dimension k of the pixels on each fragment, where k=1,... , n, n≤ N, and N is the number of pixels in the image.

3. The method according to claim 1, great for the present, however, what parts of the image select different configuration space of dimension k_ipixels on each i-th fragment, where k_i=1,... , n, n_i≤N/2, and N is the number of pixels in the image.

4. The method according to claim 1, characterized in that the image chosen partly the same and partly different in configuration space with dimension k_ipixels on eachi-th fragment.

5. The method according to claim 4, characterized in that for each piece of image k_i=k.

6. The method according to any one of claims 1 to 5, characterized in that for each image of the time sequence of images chosen by the same respective parts of the image.

7. The method according to any one of claims 1, 3 or 5, characterized in that for each image of the time sequence of images, fragments of images to choose different size and configuration, and size and configuration of the fragments chosen, taking into account the possibilities when comparing the respective current and reference image hosting and matches the configuration of the set of corresponding fragments smaller area on the corresponding piece of the big square.

8. The method according to any one of claims 1 to 3, characterized in that the pretreatment is carried out in the form of spectral filtering of images.

9. The method according to any which one of claims 1 to 3, characterized in that the reference signal is chosen characterizing magnitudes of signals of fragments of images, or, for example, each previous to subsequent, or, for example, the first image for the entire time sequence of images.

10. The method according to any one of claims 1 to 3, characterized in that the signal quality for each image slice select signals corresponding to the value obtained by the average processing signals obtained by measuring the characteristic values of the corresponding fragments for a given number of images.

11. The method according to any one of claims 1 to 3, characterized in that the reference signal selects the signal corresponding to the inverted images of the respective fragments.

12. The method according to any one of claims 1 to 3, characterized in that the parameters of the scene and the object is determined by measuring the number, order and the brightness of the fragments placed on the object and specified areas of the scene.