FR2633747A1 - Adaptive image-processing system - Google Patents

Abstract

The image to be processed is stored in a memory storage circuit 1. Modification circuits 2, 3, 7, 9 and circuits 4, 5, 6, 8, 10 for determining characteristic magnitudes of the image to be processed are arranged in cascade in order to form a single chain. A microprocessor 20 controls several successive passes of the image in the chain, setting the modification circuits 2, 3, 7, 9 and determination circuits 4, 5, 6, 8, 10 as a function of the preceding passes. The image is read in the form of a vector of pixels and processed on the move by each of the circuits. The invention applies to the detection of a target in a noisy image, for example.

Description

The subject of the present invention is an adaptive image processing system comprising - a chain of adjustable means for modifying the image to be processed, - a plurality of means for determining magnitudes characteristic of the image to be processed, and, - means for, in response to the results at the output of the determination means, adjusting the modification means so that the processing is adapted to said characteristic quantities.

Such a system can be applied to any kind of image processing, and more particularly to the determination of the position of a contrasting target in a noisy image, for example an infrared image.

Such systems are already known, in which the chain of modification means comprises a mask circuit followed by a threshold circuit, for example. The means for determining characteristic quantities include a barycenter calculation circuit, connected to the output of the chain, and a statistics calculation circuit receiving the video signal representative of the image to be processed, also received by the mask circuit. The adjustment means comprise a microprocessor arranged to, in response to the results at the output of the statistics calculation circuit, adjust the mask circuit and the threshold circuit in a manner adapted to the statistical characteristics of the image, so that the results at the output of the barycenter calculation circuit represent, for example, the coordinates of a possible target contrasted with an error as small as possible.

The main drawback of such a system is its fixed structure, which means that if, in a practical case, it is noticed that it is necessary, in order to achieve a better result, to carry out a modification of a new type, or to calculate a characteristic quantity of a new type, it is necessary to modify the whole structure of the system, which is long and clumpy. This is a major drawback, especially in a field such as image processing or, in a given situation, there are numerous trial and error before reaching a satisfactory result. In addition, when such a system breaks down, it is very difficult to locate the defective circuit due to the complex structure of the system, and the interactions of the different circuits on each other.

The present invention aims to overcome the above drawbacks.

To this end, it relates to a system of the type defined above, characterized in that - the respective video outputs of the modification means are all identical, - the respective video inputs of the modification means and of said determination means are all identical, each of them being able to receive the signal of any video output, - each of said means of determination is adjustable, and provided with a video output identical to the previous ones delivering a video signal identical to that received on its video input, - each of the determination means is inserted into said chain, - there is provided, at the input of said chain, means for storing the image to be processed, and, - the adjustment means are arranged to control several successive passages of the image to be processed in said chain, and to adjust, at each passage, the modification means and the determination means in response to the result s determined by the means of determination during the preceding passages.

In the system of the invention, the fact that the determination means are provided with a video output, and the fact that all the video inputs and all the video outputs are compatible, the structure of the system, in the form of a single chain, is very simple and allows adding, or deleting, a circuit without calling into question the architecture of the system. In addition, thanks to the storage means and the several successive passages of the image in the chain, the simplicity of the structure is without compensation for the complexity of the processing operations that can be carried out by the system.

Advantageously, the adjustment means and the storage means are arranged so that the image is read in the form of a succession of pixels, and the modification means and the determination means are provided for, during each pass, to process on the fly the succession of pixels representing the image.

The speed of the system is therefore compatible with real-time processing of moving images.

Advantageously also, a second video output is provided on each of the modification means and of the determination means, delivering a video signal identical to that of said first video output, connected to a single video bus.

So determining which elements fail in the event of a failure is particularly simple.

The present invention will be better understood from the following description of two embodiments of the system of the invention, made with reference to the accompanying drawings, in which - Figure I shows a system for the acquisition of a target by a missile; - Figure 2 shows a system for the pursuit of a target by a missile, and, - Figure 3 shows a system for the acquisition and pursuit of a target by a missile.

Referring to FIG. 1, a system for acquiring a target contrasted in an image is now described.

The image in question here is an infrared image obtained at the output of the analysis head of a seeker mounted on a missile, and the target is for example an airplane, which, because of its temperature very much higher than the background of the sky on which it is located, appears, in a contrasting manner on the infrared image.

By acquisition is meant the determination of the position of the target in the image, it being understood that, a priori, the position of this target is arbitrary. As is well known to those skilled in the art, the acquisition phase is the phase during which the missile searches for its target. To this end, the infrared analysis head is controlled to scan the widest possible field. As soon as the acquisition has taken place, that is to say as soon as the target has been detected and its position determined, the missile goes into the pursuit phase, during which the control surfaces are controlled so that the target appears at center of the field and remains there, the amplitude of the scanning of the analysis head then being substantially reduced. A tracking system will be described with reference to Figure 2.

 The acquisition system of FIG. 1 comprises, arranged in cascade at the output of the analysis head, not shown because it is conventional, a storage circuit 1, a circuit 4 for detecting local maxima, a circuit 5 for calculating statistics , a filtering circuit 7, a circuit 9 for comparing the background noise, and a circuit 10 for storing alarms.

The storage circuit 1 is provided with a digital video input on which it receives, from the analysis head, a digital video signal IO, supported by a bus 50, and comprising a series of words by 12 bits , associated with clock, synchronization and validation signals known to those skilled in the art. Each of these words represents a pixel of the infrared image, and results from the analog-digital conversion of the analog signal representing the intensity of the infrared radiation measured for this pixel, possibly corrected to take account of the defects of the head. analysis.

The storage circuit 1 is provided with a digital video output which delivers a signal II, comprising, like the signal IO, a series of words of 12 bits each representing a pixel, associated with a clock signal, a validation signal and a synchronization signal. The signal It is supported by a bus 51, which here comprises 19 conductors distributed as follows: one for each of the 12 bits representative of the pixel, one for the clock signal, one for the validation signal, one for the synchronization signal , and four not used.

The local maximum detection circuit 4 is provided with a digital video input, receiving the signal Il ', and with a digital video output delivering a signal I4, supported by a bus 54 identical to the bus 51.

Similarly, the statistics calculation circuit 5, the filtering circuit 7, and the background noise comparison circuit 9 are each provided with a digital video input receiving respectively the signal I4, a signal I5 and a signal I7, and each of a digital video output respectively delivering the signals I5, I7, and a signal I9. Signals I5, I7, and I9 are supported respectively by buses 55, 57 and 59, identical to buses 51 and 54.

Finally, the alarm storage circuit 10 is provided with a digital video input receiving the signal I9.

Thus, the digital video outputs of circuits 1, 4, 5, 7 and 9 are all identical, as well as the digital video inputs of circuits 4, 5, 7, 9 and 10, and each of these inputs could therefore receive the signal delivered by each of these outputs.

As shown in FIG. 1, each of the circuits 1, 4, 5, 7, 9 and 10 is provided with an access connected to a microprocessor 20 via a bidirectional bus 60.

As will be better understood hereinafter, each circuit 1, 4, 5, 7, 9 and 10 acts on the image which it receives either to modify it or to determine one or more quantities characteristic of this image. To this end, each of these circuits receives on the access which connects it to the bus 60, reconfiguration data, that is to say adjustment of the tasks which it must perform on the video signal which passes through it. When these tasks are performed, data representing the result or results of these tasks are available on the same access, where they are read by the microprocessor 20.

Each of the circuits 1, 4, 5, 7, 9 and 10 is provided with an additional digital video output, delivering a signal identical to that delivered on the digital video output previously defined, connected to a bus 70. As will be better understood in the following, this bus, called test bus, is not used in normal operation, but allows in the event of a malfunction of the system, for example, to determine which circuit is faulty.

Thus it appears that circuits 1, 4, 5, 7, 9 and 10 are arranged so as to form a single chain, in which can pass, in the form of signals Il, I4, 15, I7, and I9, successively, a image to be processed stored in the storage circuit 1, in order to be modified therein and so that certain of the quantities characteristic of this image are determined, under the orders of the microprocessor 20.

Before tackling the detailed operation of the system, each of the circuits 1, 4, 5, 7, 9 and 10 is now described.

The main function of the storage circuit 1 is the storage of the infrared image at the output of the analysis head. In addition, it allows the change of rhythm between the writing of the digitized pixels of the signal IO, and the reading of those of the signal Il. Indeed, the writing rate here is substantially 1 ps per pixel, while, as will be better understood below, the reading rate is significantly ten times faster, that is to say 100 ns per pixel.

Furthermore, the pixels of the image to be memorized being organized, in a known manner, in rows and columns, the memorization circuit allows the writing and the reading of the pixels, as desired, row by row, or column by column, from right to left or vice versa, and from top to bottom or vice versa. This characteristic makes it possible, for example, to adapt the directions of reading or writing to the direction of scanning of the analysis head, direction of scanning which is often reversed from one image to the next. In this way, the signal Il always corresponds to the same arrangement of the target image, and not to an inverted arrangement from one image to the next, which saves time on subsequent processing. The choice of types and direction of reading or writing is made by the microprocessor 20, via the bus 60.

Finally, the storage circuit 1 allows the reading of the pixels in only a portion of the stored image, defined portion rar the rank of its first row and of its first column and the number of its rows and of its columns, and called in the continuation "sub-image". This will, as will be better understood hereinafter, save time for certain treatments. The choice of the sub-image is made by the microprocessor 20, via the bus 60.

The storage circuit 1 here comprises two identical RAM type memories, each allowing the storage of an image of 512 x 512 pixels, and all the circuits necessary for the addressing of these memories, such as this addressing has just been described. .As the pixels of the Io signal arrive continuously, and we want to carry out, in the channe, a processing in real time as fast as possible, each of the two memories is alternately used in reading and writing, the IO signal being written in one while the other is read to generate the signal Il. The design of a circuit comprising the two memories above and addressing circuits, arranged to operate as just described, is within the reach of the skilled person and circuit 1 will therefore not be more described.

 The local maxima detection circuit 4 is arranged to compare the intensity of each pixel of the signal II with the intensity of each of the eight pixels closest to this pixel on the infrared image, and to deliver a signal I4 identical to siganl II, except for a delay time. Circuit 4 calculates each of the eight intensity differences between this pixel and its eight neighbors. If these eight differences are all greater than a threshold, the pixel is considered to be a local maximum. The threshold value is adjusted by the microprocessor 20, via the bus 60. The coordinates of this local maximum are quantities characteristic of the infrared image, determined by the circuit 4-of detection of local maxima. These coordinates can be transmitted to the microprocessor 20 so that it stores them for later use. However, since four conductors of the bus 54 are unused, it is also possible, and equivalent, to use one of these conductors as a local maximum indicator, to transmit the results determined by the circuit 4 downstream of the latter.

For this purpose, the circuit 4 is arranged so that, at its output, when the levels of the 12 data conductors represent the 12 bits of a pixel, the level of the local maximum indicator conductor is high, for example, if this pixel is a local maximum, and low otherwise.

It will be noted that circuit 4 having to compare the value of the intensity of a pixel with that of its eight neighbors to determine if this pixel is or is not a local maximum, it is necessary, in the case where one uses a local maximum indicator, as has just been described, ## a certain time elapses between the time of entry of the information relating to a pixel in the circuit 4, and their time of exit. This delay between the signal I4 and the signal It is at least equal to the period of time necessary for the information relating to each of the eight pixels neighboring the processed pixel to be known to the circuit 4.

It will be noted that the circuit 4 does not modify the image which passes through it, since the signal I4 is identical to the signal Il, except for the delay which has been mentioned. However, as will be better understood below, there are situations where there is no need to detect the local maxima. In this case, it is useless to waste time because of the passage in the circuit 4. There is therefore provided, upstream of the video output of the circuit 4, a multiplexer with two inputs, one connected to the video output of the working circuits of circuit 4, the other connected directly to the video input buffer of circuit 4. The microprocessor 20 controls this multiplexer, via bus 60, so that the video signal passes through the through the working circuits, or directly to the output. In the second case, it is said that the microprocessor 20 regulates the circuit 4 so that it is "transparent". This characteristic is not particular to circuit 4 and it is systematically provided for on all circuits which will be described hereinafter as being provided with a video input and a video output, and adjustable by the microprocessor 20. For reasons simplicity, so nty will no longer be alluded to.

The design of a calculation circuit carrying out the functions which have just been described with regard to circuit 4 for detecting local maxima is within the reach of a person skilled in the art, and this circuit will therefore not be described further.

The circuit 5 for calculating statistics is arranged to determine magnitudes characteristic of the image represented by the pixels of the signal I4 and to deliver a signal I5 identical to the signal I4, with a delay time. These quantities are here as follows - number of pixels of zero intensity, - number of pixels of non-zero intensity, - maximum intensity of the image and coordinates of the pixel (s) of maximum intensity, - minimum intensity of the image and coordinates of the pixel (s) of minimum intensity, - average intensity of the image, - "effective" intensity or RMS of the image, and, - quadratic mean value of the intensity.

The choice of the quantity or quantities to be determined by the circuit 5 is made by the microprocessor 20, via the bus 60. Likewise, the results determined by the circuit 5 are read by the microprocessor 20 thanks to the bus 60. The design of a calculation circuit carrying out the functions which have just been described in connection with circuit 5 is within the reach of a skilled person, and this circuit will therefore not be described further.

The filter circuit 7 is arranged to filter the image represented by the signal pixels
I5, the filtered image being represented by the pixels of the signal I7. The filtering can be either high-pass, low-pass, or a 3 x 3 Sobel-type convolution, to bring out the horizontal lines. The choice of the type and parameters of the filtering 7 is made by the microprocessor 20, via the bus 60. As the filtering circuit 7 modifies the image represented by the pixels of the signal I5, without determining the quantities characteristic of this image, the access to circuit 7 connected to bus 60 is unidirectional, that is to say that it only supports information entering circuit 7.

We can say that this information, relating to the type and filtering parameters carried out by the circuit 7, relates to the adjustment of the circuit 7 by the microprocessor 20. The design of a calculation circuit carrying out the functions which have just been described about circuit 7 is within the reach of the skilled person, and this circuit will therefore not be described further.

The background noise comparison circuit 9 is arranged to compare the intensity of each pixel with the local background noise around this pixel, and to replace it with a pixel of zero intensity when it is insufficiently contrasted with respect to the noise. background, or when, sufficiently contrasted, it was not detected as local maximum by circuit 4. When on the other hand, a pixel is sufficiently contrasted and when, moreover, it was detected as local maximum by circuit 4 , the pixel is replaced by a pixel of maximum intensity. The image thus modified by circuit 9 is represented by signal I9. More precisely, the local background noise around a pixel is calculated by circuit 9 as being the local average over the square of 3 x 3 or 5 x 5 pixels, for example, at the center of which this pixel is located. pixel is considered to be ~ sufficiently contrasted if the difference between the pixel intensity and the local background noise is greater than a threshold. If this is the case, and if the microprocessor 20 or the signal on the conductor indicating the local maximum indicates that the pixel is a local maximum, the pixel is replaced by a pixel of maximum intensity.

 The choice of the size of the square used for the calculation of the local background noise, as well as that of the threshold, for example, are made by the microprocessor 20, via the bus 60. Here again, as the circuit 9 modifies the image without calculating a characteristic magnitude thereof, no result is transmitted from circuit 9 to microprocessor 20, and access from circuit 9 connected to bus 60 is unidirectional to circuit 9. The design of a circuit computation carrying out the functions which have just been described with regard to circuit 9 is within the reach of those skilled in the art and this circuit will therefore not be described further.

The alarm storage circuit 10 is arranged to determine and store, if necessary, the position of the pixel of maximum level in the image represented by the pixels of the signal I9, a pixel that is considered to be an alarm, ctest- ie as a potential target, since, as has been explained, this pixel is a local maximum, and it is well contrasted with respect to the local background noise. The circuit 10 essentially comprises a circuit for determining the position of the alarm followed by a memory of the FIFO type ("First Input First Output", in which the position of the alarm is stored). alarm which is a characteristic quantity of the image represented by the signal I9. The microprocessor 20 dialogues with the circuit 10 via the bus 60. The design of a circuit carrying out the functions which have just been described in connection circuit 10 is within the reach of ordinary skill in the art and this circuit will therefore not be further described.

The acquisition system of FIG. 1, the structure of which has just been described, operates as follows.

The microprocessor 20 is arranged to control several successive passages, or "flows", in the chain comprising circuits 4, 5, 7, 9 and 10, of the image stored in circuit 1, that is to say of the image to be processed.

During each pass, the image stored in circuit 1 is first read, and the corresponding video signal passes successively through each of circuits 4, 5, 7, 9 and 10. Between one pass and the next, the image stored in circuit 1 is unchanged, but the settings of circuits 4, 5, 7, 9 and 10 are possibly modified During a passage, the settings depend on the results relating to the characteristic quantities of the image which have been determined during previous passages, that is to say that the system is adaptive.

Thus, and by way of example, there are provided, in the acquisition system which has just been described, three successive passages.

Before the first pass, the microprocessor 20 controls, or regulates, the circuit 5 so that it calculates the maximum intensity, the minimum intensity, the average intensity, and the effective intensity, these characteristic statistical quantities being intended to be used during the third pass. Still before the first pass, the microprocessor 20 adjusts the filter circuit 7 so that it is configured as a high-pass filter, and in particular adjusts the threshold of the circuit 9 for comparison with background noise, to a predetermined value.

The microprocessor 20 controls. then the successive reading of each of the pixels of the image stored in circuit 1, to effect the first pass, or first flow.

When the first pass has taken place, the microprocessor 20 reads the results obtained in particular at the output of circuits 5 and 10 and stores them. It is clear that the first pass here made it possible to detect a possible point target, that is to say a target whose image is substantially the size of a pixel.

Before the second pass, and in the particular case described here, the microprocessor 20 does not change the settings of the circuits of the chain.

However, for the second pass, the microprocessor controls the successive reading of only one pixel out of two of the stored image. The second pass is therefore identical to the first, except that the image being undersampled, this second pass allows the detection of a possible average target, that is to say of a target whose image has roughly the size of a 2x2 pixel square.

Between the second and the third passage, the microprocessor 20 reads the results obtained in particular at the output of circuits 5 and 10, stores them and changes the settings of circuits 4, 5, 7 and 9. Circuit 4 for detecting local maxima and the circuit 5 for calculating the statistics are made transparent. The filtering circuit 7 is set to be configured as a low-pass filter, and the threshold of circuit 9, for comparison with background noise, is adjusted according to the statistical results determined during the first pass.

During the third pass, the microprocessor controls the successive reading of each of the pixels of the image stored in circuit 1, which allows the determination of a possible wide target, that is to say of a target whose the image is roughly the size of a 25 pixel square, for example.

After these three successive passes, If no target has been found, the microprocessor 20 controls three new passes, and so on, until a target is found. The control surfaces of the missile are then controlled so that the missile heads towards the target, which has the effect of bringing the image of the target to the center of the image.

The acquisition phase is then completed, and the missile enters the pursuit phase.

Referring now to FIG. 2, the tracking system, which can therefore be used when a target has been acquired using the acquisition system of FIG. 1, comprises, arranged in cascade at the output of the analysis head, a storage circuit 1 identical to that already described, a circuit 2 forming a mask, a circuit 3 forming a threshold, a circuit 5 for calculating statistics identical to that already described, a circuit 6 for calculating barycenters, a filtering circuit 7 identical to that already described, and a circuit 8 jet detector.

 As before, the storage circuit 1 delivers the signal Il on the bus 51. The circuits 2, 3, 5, 6 and 7 are each provided with a digital video output identical to those already described, delivering signals I2, I3, I5, I6 and I7 respectively, supported by buses 52, 53, 55, 56 and 57 respectively, these buses being all identical to bus 51.

Likewise circuits 2, 3, 5, 6, 7 and 8 are each provided with a digital video input receiving the signals Il, I2,
I3, I5, I6 and I7 respectively. All digital video inputs are identical to those already described.

As before, each of the circuits 2, 3, 5, 6, 7 and 8 is provided with an access connected to a microprocessor 20, identical to that already described, via a bidirectional bus 60 ', having the same bus 60. Similarly, each of the circuits 2, 3, 5, 6, 7 and 8 is provided with an additional digital video output delivering a signal identical to that delivered on the digital video output previously defined, and connected to a test bus 70 'having the same functions as bus 70.

Here again, it appears that circuits 2, 3, 5, 6, 7 and 8 are arranged so as to form a single chain through which signals Il, I2, I3, can pass,
I5, I6, I7 and I8, successively, an image to be processed stored in the storage circuit 1, to be modified there, and so that certain of the quantities characteristic of this image are determined, under the orders of the microprocessor 20.

Before tackling the operation of the system, each of the circuits 2, 3, 5, 6, 7 and 8 is now described.

The mask forming circuit 2 is arranged to mask either the inside or the outside of a certain number of windows, here rectangular and the number of 3, each window being of adjustable position and dimensions. Circuit 2 delivers a signal I2 in which the masked pixels have zero intensity, the non-masked pixels having a value identical to that which they present in the signal Il. Circuit 2 is therefore an image modification circuit. The choice of the position and dimensions of each window, defined by the rank of its first row and its first column, and the number of its rows and columns, as well as the type of masking, inside or outside each window , is made by the microprocessor 20, via the bus 60 '.

Access to circuit 2 connected to bus 60 'is unidirec tional to circuit 2. The design of a computing circuit realizing the functions which have just been described in connection with circuit 2 is within the reach of the skilled person. profession, and this circuit will therefore not be described further.

The threshold circuit 3 is arranged to modify some of the pixels of the image represented by the signal I2, and deliver a signal 13 representing the modified image, using one of the following criteria - modification of the pixels of intensity less than a low threshold, to make their intensity zero, for example, - modification of the pixels of intensity greater than a high threshold, to make their maximum intensity, for example, and, - modification of the pixels of intensity between a high threshold and a low threshold, to make their intensity uniform, for example.

In addition, circuit 3 determines the following quantities, characteristics of the image - number of pixels of intensity below the low threshold, - number of pixels of intensity above the high threshold, and, - number of pixels of intensity included between the high threshold and the low threshold.

The microprocessor 20 notably regulates the value of the high and low thresholds, and collects the quantities determined by the circuit 3 via the bus 60 '. The design of a calculation circuit carrying out the functions which have just been described in connection with circuit 2 is within the reach of a person skilled in the art, and this circuit will not be described further.

The circuit 6 for calculating the barycenters is arranged to calculate the "thermal barycenter" and the "geometric barycenter" of the image represented by the signal I5. The "thermal barycenter" is the barycenter of the image, in the classic sense of the term. The "geometric barycenter" is the barycenter of the image obtained after the intensity of all the pixels of intensity less than a threshold is canceled, that of all the other pixels of the image being made maximum. In such an image, the target appears, if necessary, as a white spot on a black background, and the position of the barycenter of this spot appears to be linked to the geometry of this spot. The circuit 5 delivers a signal I6 identical to the signal I5, except for a delay. The microprocessor 20 adjusts the threshold value, controls the calculation of the barycenters and reads their coordinates via the bus 60 '. Such a circuit is within the reach of a skilled person, and it will not be described further.

Circuit 8, a jet detector, is arranged to search, on each horizontal line of the image represented by signal I7, for the longest series of successive pixels of intensity greater than a threshold, as well as to determine the coordinates of the first pixel of the sequence, and the number of pixels of the sequence. The microprocessor 20, via the bus 60 ′, sets the threshold value and reads the characteristic magnitudes of the image that are for each line the coordinates and the number which have just been defined. The design of a calculation circuit carrying out the functions which have just been described in connection with circuit 8 is within the reach of a person skilled in the art, and this circuit will therefore not be described further.

The tracking system, the structure of which has just been described, operates, for example, in the following manner.

As with the acquisition system, the microprocessor 20 controls several successive passages of the image memorized in the chain, by adjusting, during each passage, each of the circuits of the channel in response to the results determined during the preceding passages.

Remember that, before the first pass, the target is centered, at least approximately, and that, during the acquisition procedure, it has been determined whether it is a point target, medium or wide . The microprocessor 20 then adjusts. circuit 2 so that the interior of a centered window is hidden, of a size adapted to the size of the target which has been determined.

 Thus, depending on whether, in the tracking phase, a point, medium or large target has been detected, the microprocessor 20 chooses one of the windows from a set of three, each of which is of predetermined size to mask, for sure. of a target of the corresponding size. This window will be called in the following "target window".

The microprocessor 20 adjusts circuits 3, 6, 7 and 8 so that they are transparent, and it controls, or adjusts, circuit 5, for calculating statistics, to calculate the average intensity and the effective intensity of l image represented by the signal I3.

The microprocessor 20 then controls the successive reading of each of the pixels of the image stored in the circuit 1, to effect the first pass, or first flow. During this first flow, the only active circuits are circuit 2 and circuit 5. Because circuit 2 masks the target window, circuit 5 operates on the unmasked part of the image. This part is here annular around the target window, and will be designated in the following as the "landscape window".

After the first pass, the results of the statistical calculations made by the circuit 5 on the landscape window are read by the microprocessor 20, which deduces a low threshold which will be used in a later pass. The microprocessor 20 then changes the settings for the second pass. Here, it simply modifies the setting of circuit 2 forming a mask, to make it transparent.

The microprocessor 20 then controls the second pass.

For this, it controls the storage circuit 1 so that only the pixels of a sub-image coinciding with the target window are read. Thus, circuit 5 operates only on the interior of the target window.

After the second pass, the results of the statistical calculations performed by the circuit 5 on the target window are read by the microprocessor 20, which deduces a high threshold which will be used in a later pass.

It will be noted that, to achieve the same result during the second pass, the microprocessor 20 could command circuit 1 to read the entire image, and circuit 2 to mask the outside of the target window, i.e. - make the landscape window. However, since, as is the case here, it is possible to command the reading of a sub-image in the memory of circuit 1, this solution has the advantage of being faster to achieve the same result. .

After determining the high threshold, the microprocessor 20 then changes the settings for the third pass. Circuit 2, forming a mask, is set to be transparent.

Circuit 3, forming a threshold, is adjusted to modify the image by canceling the intensity of the pixels of intensity below the low threshold determined during the first pass, and by maximizing the intensity of the pixels of intensity above the threshold. high determined during the second pass. Circuit 5, for calculating statistics, is set to be transparent. Circuit 6 is set to be active, i.e. to calculate the thermal barycenter and the geometric barycenter. Circuit 7 is set to bring out the horizontal lines. Finally, circuit 8 is set to be active, that is to say to detect the possible hot jet of a reaction propulsion system with which the target can be fitted. It will be noted that a device is provided for stabilization of the missile with respect to the vertical, and that such a jet must in principle appear to be horizontal.

When these adjustments are made, the microprocessor 20 controls the third pass, still only reading, in circuit 1, the pixels inside the target window.

After the third pass, the microprocessor 20 reads the results from the circuit 8, jet detector, and interprets them as follows.

If a jet was detected during the third pass, the microprocessor 20 controls a fourth pass, for which the circuits of the chain are in the same configuration as previously, but for which the microprocessor 20 reduces the size of the target window, it that is to say here of the sub-image read in the storage circuit 1, so that the jet is not taken into account in this window. The target window thus reduced is called the airplane window, insofar as it corresponds only, in principle, to the structure of the airplane, and not to the assembly, of the structure and of the hot jet of the propulsion system. At the end of the fourth pass, the microprocessor 20 reads the coordinates of the geometric barycenter of the airplane window, which will then be considered as the coordinates of the position of the target in the image.

If no jet was detected during the third pass, the microprocessor 20 assesses the distance of the target from the size of its image. If the target is far away, the microprocessor 20 will read the coordinates of the thermal barycenter determined by the circuit 6, which will then be considered as the coordinates of the position of the target in the image. If the target is close, the microprocessor 20 will read the coordinates of the geometric barycenter determined by the circuit 6, which will then be considered as the coordinates of the position of the target in the image.

Thus, in three or, if necessary, four passes of the image in the chain, the coordinates of the position of the target are determined and can be transmitted to the control system of the control surfaces of the missile to pursue the target. The microprocessor 20 will then process the next image, and so on. Naturally, for the processing of an image of given rank, the microprocessor 20 uses the information which could have been determined by virtue of the processing of the preceding images.

In particular, the microprocessor 20 adjusts, for each image, the dimensions of the target window so as to best follow the evolution of the apparent size of the target, by comparing the intensity of the pixels of the image at the low threshold determined at during the processing of the previous image. By processing an image, it is naturally necessary to understand all of the three or four passages of this image in the chain, as just described.

Naturally, on a missile, it is not necessary to provide two separate chains such as the chains of FIGS. 1 and 2 which, for the sake of clarity, have been described separately. Particularly, as shown in Figure 3, one can advantageously use a single chain comprising for example circuits 1 to 10 arranged in cascade, and, for example, in this order. In this case, the jet detector circuit 8 is provided with a digital video output identical to the others.

Then, bus 60 is connected to each of circuits 1 to 10.

For acquisition, the microprocessor 20 makes circuits 2, 3, 6 and 8 transparent. For tracking, it makes circuits 4, 9 and 10 transparent.

 In the chains which have been described here, the clock signal, the period of which is approximately equal to the time necessary to read a pixel in the storage circuit 1, that is to say 100 ns in the example described, is supported by one of the conductors of buses 51 to 59. We can say that the clock is mixed with the data, and delayed with it in circuits I to 10, like the synchronization signal, which includes a pulse before each data stream , and the validation signal, at the high level as long as the transmitted data are valid.

The data changes approximately at each rising edge of the clock signal, and their value is determined at each falling edge.

Each circuit 2 to 10 operates synchronously. The data as well as the synchronization and validation signals are resynchronized at the input of the circuit, by the input clock signal. Each circuit delivers to the following circuit a synchronous clock signal of the data and of the synchronization and validation signals. With each clock stroke, the data shifts and thus traverses all the circuits assembled in the chain.

The solution which consists in mixing the clock signal with the data is simple, but can lead to difficulties if too many circuits are connected in cascade, in particular because the rise times of the various components differ from their fall times. To avoid these difficulties, it is of course possible to use the same clock signal applied simultaneously to each of the circuits 1 to 10.

The total duration of a passage, or flow, in the chain, between the moment when the first pixel of the image leaves circuit 1 and the moment when the last pixel of image is processed by circuit 10, for example, is only very little longer than the time required to read the image in circuit 1, since the delay time introduced by each of circuits 1 to 10 between its video input and its video output is only a few clock strokes.

Thus, and to fix ideas, when the reading time of a pixel in circuit 1 is of the order of 100 ns, an image of 32 x 32, or 1024 pixels, requires a total reading time of substantially 100 ps.

The clock period being 100 ns, the time necessary for a pixel to pass from one end to the other of the chain is a few PE, for example 2 # 8. Thus, the first pixel of the image which leaves circuit 1 reaches the end of the chain, for example at the exit from circuit 10, at the moment when the twentieth pixel of the image leaves circuit 1. Thus the duration of a flow, here of 102 They appear as mainly linked to the image reading time, here 100 ps.

We see that, in the tracking phase, when processing 32 x 32 images, or 1024 pixels, the total processing time for an image is a few milliseconds. Indeed, with a writing time of about 1 iis per pixel, it takes first of all a duration of about 1 ms to write 11 images in the memory circuit 1. Then, for each of the passages, a slightly longer duration at 100 s so that all the pixels of the image pass once through the chain, and a duration of around 1 ms so that the microprocessor reads the determination results and proceeds to change the setting for the next pass.

A total image processing time of a few ms is thus obtained, which is compatible with real-time processing, which corresponds to a frame rate of one image every 10 ms when the target is close, and one every 40 ms when it is far away, because then its image moves rather slowly.

In the acquisition phase, this duration of a few milliseconds may be slightly increased, in particular because the image processed can be larger. Indeed, if one uses, for example, an analysis head comprising a movable mirror in front of a mosaic of 32 x 32 detectors, the scanning of the mirror in the acquisition phase is controlled to form images in the form of a strip of 32 x 256 pixels, for example. In the pursuit phase, on the other hand, the mirror is left stationary, and images of 32 x 32 pixels are formed. The writing time of a strip is 200 m s for the acquisition, because this time is linked to the scanning speed of the mirror. On the other hand, it takes approximately 800 p s to read a banner, which slightly increases the processing time.

We can say that the chain in Figure 3, for example, appears as the cascading of circuits each processing, on the fly, an image in the form of a vector, or a succession, of pixels, here in number of 1024 or 8192, since the processing of the first pixels is finished a very long time before that of the last ones begins.

As already mentioned, the role of buses 70 to 70 'is to allow the diagnosis of the origin of a system failure, step by step, without having to dismantle it. To do this, the bus 70 or the bus 70 ′ is connected to a video monitor, and a typical image, for example a test pattern, of known characteristics, is loaded into the storage circuit. Step by step, we can thus control the operation of each of circuits 2 to 10, by examining, on the monitor, the image of the target, modified or not, at the output of each of circuits 2 to 10, or by comparing the results of the determinations of the characteristic quantities, read by the microprocessor 20 in these circuits, with what they must be for the test chart.

In this case, the possibility of making any of the circuits 2 to 10 transparent is obviously of great interest. The failure can thus be detected without difficulty and the faulty circuit 2 to 10 quickly replaced by another circuit functioning properly.

Naturally, the invention is not limited neither to the two methods of implementing the systems of FIGS. 1, 2 or 3 which have been described, nor even to the specific structures of these systems, structures which have only been described as an example. Indeed, it is obviously within the reach of a skilled person to develop other methods of using the systems described, or even to add or remove one or more circuits to these systems, so as to perform image processing specific to the particular problem it may have to solve.

In the system of the invention, it is the cascading, to form a single chain, of image modification circuits, and of circuits for determining magnitudes characteristic of these images, and the adaptive adjustment of these circuits to the during a sequence of several successive passages, on the fly, of the image in the chain, which makes it possible to achieve the result, namely a system for processing animated images in real time, of modular structure, in which the different circuits or modules are easily interchangeable, this system being capable of admitting new circuits or modules if necessary, and this system being easy to troubleshoot.

 In fact, the acquisition and tracking systems which have been described appear as adaptive image processing systems, in which the image is processed in a manner adapted to the quantities characteristic of this image, such as its average intensity, its effective intensity, the presence or absence of a jet, and so on. It can be said that the image to be processed is the noisy image of a contrasted target, the result of the processing being the coordinates of the target in the image.

Claims (6)

1.- Adaptive image processing system comprising - a chain of adjustable means (2, 3, 7, 9), for modifying the image to be processed, - a plurality of means (4, 5, 6, 8, 10) for determining characteristic quantities of the image to be processed, and, - means (20) for, in response to the results at the output of the determination means (4, 5, 6, 8, 10), adjusting the means of modification (2, 3, 7, 9) so that the processing is adapted to said characteristic quantities, system characterized in that - the respective video outputs of the modification means (2, 3, 7, 9) are all identical, - the respective video inputs of the modification means (2, 3, 7, 9) and of the determination means (4, 5, 6, 8, 10) are all identical, each of them being able to receive the signal of an output any video, - each of said means of determination (4, 5, 6, 8, 10) is adjustable and provided with a video output identical to the previous deliveries t a video signal identical to that received on its video input, - each of the determination means (4, 5, 6, 8, 10) is inserted in said chain, - there is provided, at the input of said chain, means (1) for storing the image to be processed, and, - the adjustment means (20) are arranged to control several successive passages of the image to be treated in said chain, and to adjust, with each passage, the means of modification (2, 3, 7, 9) them means of determination (4, 5, 6, 8, 10), in response to the results determined by the means of determination (4, 5, 6, 8, 10) during previous passages. 2. The system as claimed in claim 1, in which the adjustment means (20) and the storage means (1) are arranged so that the image is read in the form of a succession of pixels, and the modification means ( 2, 3, 7, 9) and the determination means (4, 5, 6, 8, 10) are provided for, during each pass, to process on the fly the succession of pixels representing the image. 3.-System according to one of claims 1 or 2, wherein the image to be processed is the noisy image of a contrasting target, and the result of the processing is the position of the target in the image. 4.- System according to one of claims 1 to 3, wherein there is provided on each of the modification means (, 2, 3, 7, 9) and determination means (4, 5, 6, 8, 10 ) a second video output #, delivering a video signal identical to that on said first video output, connected to a single video bus (70; 70 ').  5.- System according to one, of claims 1 to 4, in which the modification means comprise filtering means (7) and means (9) for comparison to background noise, and the determination means comprise means (4) for detecting local maxima, means (5) for calculating statistics and means (10) for storing alarms. 6.- System according to one of claims 1 to 4, wherein the modification means comprise mask means (2), threshold means (3), filtering means (7) and means (8) jet detectors, and the means for determining comprise means (5) for calculating statistics and means (6) for calculating barycenters.