Short-term spectral event detection

Abstract

A system for the detection of a short-term spectral burst event, the system comprising a processing module configured for receiving raw RGB data in the form of a set of image frames obtained by an image capturing device which comprises at least one sensor comprising R, G, and B channels; the processing module is further configured for filtering the set of image frames by removing non-varying pixels between each two consecutive image frames of at least a portion of the set of image frames, thereby extracting a first filtered pixel set comprising a reduced amount of data compared to the set of image frames; the processing module is also configured for performing a spectral analysis of the R, G and B channels of at least a portion of the first filtered pixel set based on a spectral ratio between properties of the R channel and properties of the G and B channels, over time, and, based on the spectral analysis, determining the existence of the short-term spectral burst event.

Description

[0001] SHORT-TERM SPECTRAL EVENT DETECTION

[0002] TECHNOLOGICAL FIELD

[0003] The present invention is in the field of short-term spectral even detection, in particular, the detection of launch, explosion or fire from a firearm.

[0004] BACKGROUND OF THE INVENTION

[0005] It is well known in the art to monitor the environment using optical sensors in order to detect the presence of forest fires and similar events by performing an RGB analysis of images captured by the optical sensors.

[0006] These methods are based on various calculations and comparisons of the RGB channels, their intensity, hue, brightness level etc., and are configured for continuous monitoring of evolving fires over a significant duration of time.

[0007] Some examples are described in US2015 / 0242687, US2010 / 0034420, US2015 / 0169984, and US2003 / 0044042.

[0008] Acknowledgement of the above references herein is not to be inferred as meaning that these hinder, in any way, the patentability of the presently disclosed subject matter.

[0009] GENERAL DESCRIPTION

[0010] In accordance with one aspect of the subject matter of the present application, there is provided a system for the detection of a spectral event, said system comprising a processing module configured for:

[0011] - Receiving raw RGB data in the form of a set of image frames obtained by an image capturing device with at least one sensor comprising R, G, and B channels;

[0012] - Filtering said set of image frames by removing non-varying pixels between each two consecutive image frames of at least a portion of said set of image frames, thereby extracting a first filtered pixel set; Analyzing the R, G and B channels of at least a portion of said filtered pixel set; and

[0013] - Based on the spectral analysis of the R, G and B channels, determining the existence of said spectral event.

[0014] The spectral analysis performed by the processing module may be configured for detection of various types of spectral events, most notably, a ‘short-term spectral burst event’ and a ‘long-term spectral event’.

[0015] Such a spectral event may occur as a result of, but not limited to, a burst of fire, launch of a missile or rocket, fire from a firearm, explosion etc.

[0016] The term ‘short-term spectral burst event’ shall be used herein to define a transient spectral event occurring over a brief period of time, and having a sharp spectral rise and a similarly sharp spectral decline in the RGB channels. In particular, such an event may range from 1msec or sub 1msec (for small firearms such as handguns) all the way to 60msec (for higher ordnance explosion events). It should thus be understood that the short-term spectral burst event of the present application is in the time range of 25-30hz, reflected in a range of 1-3 frames on regular video capture, whereby minimal manageable information is given for target signature evolution or pattern in time.

[0017] Similarly, the term ‘long-term spectral event’ shall be used herein in to define a steady state event, occurring over an extended period of time, and having a mild spectral rise and either a mild spectral decline or no spectral decline at all. Such a spectral event may occur as a result of a forest fire, sunset and other prolonged events.

[0018] The system may further comprise an image capturing device with at least one sensor comprising R, G, and B channels, and configured for obtaining said raw RGB data and outputting it to said processing module.

[0019] The image capturing device may be directed to an area of interest which is to be monitored by the system. In accordance with one example, the system may comprise an image capturing arrangement comprising one or more such image capturing devices, configured for coverage of a wide area of interest. Specifically, each image capturing device may define a field of view determining the coverage of a portion of the area of interest for that specific image monitoring device. The arrangement may be such that the combined areas of interest of each image capturing device are sufficient for covering the desired area of interest of the entire system.

[0020] The system may be static, in the sense that each image capturing device is directed to a fixed location and configured for monitoring a given area of interest. In this case, the individual areas of interest of the one or more image capturing device may overlap in order to provide full coverage. Alternatively, or additionally, the system may also be configured for moving and / or rotating, such that each of the one or more image capturing devices browses over the entire area of interest of the system, each time capturing images of a different portion of the entire area of interest.

[0021] Each image frame obtained by the one or more image capturing devices may be represented by a set of pixels, each pixel recording data from a given point of the observed area of interest. For a set of image frames acquired, when no event occurs within the area of interest, a given pixel will provide very similar if not identical data between subsequent image frames. However, whenever an event causes a change in the area of interest, the pixels of an image frame directed to that event will records a change between consecutive image frames.

[0022] For the sake of clarity, the terms ‘varying pixels’ and ‘non-varying pixels’ will be used herein to respectively denote pixels which change between two consecutive image frames, and pixels which remain the same between two consecutive image frames.

[0023] The system of the present invention may be configured for monitoring the changes in the pixels of the image frames, filtering them and then analyzing the changed pixels in order to determine the occurrence of a desired event. More particularly, the system may be configured for monitoring the individual changes of pixels directed to one or more given points of the observed area. Specifically, pixels which do not exhibit any change between consecutive image frames acquired by the one or more image capturing devices are of lesser interest, and may therefore be given a lower weight or, alternatively, be completely ignored when analyzing the image frames.

[0024] This approach allows drastically minimizing the amount of data which needs to be analyzed by the processor in order to determine the occurrence of a spectral event. Specifically, instead of analyzing all of the pixels of an image, only a filtered set containing a select number of pixels is analyzed, reducing computing power, costs, latency etc. For this purpose, the processing module may comprise a filtering module, configured for extracting the filtered set of varying pixels, and an analysis module configured for performing the RGB analysis on the filtered set of pixels.

[0025] Since the changes in the filtered set of pixels may be indicative of various events taking place within the observed area of interest (e.g. moving objects, changes in the scenery, color changes due to time of day or even long-term spectral events like a sunset or a fire), the analysis module relies on an analysis algorithm configured for determining which type of spectral event the filtered set of pixels is indicative of.

[0026] For example, in order to determine if the filtered set of pixels is indicative of a short-term spectral burst event, the R, G, and B channels of the filtered set of pixels may be compared to each other and over time. A short-term spectral burst event is likely to exhibit a significant change in one of the channels compared to the others, and which is more, this change will likely take place over a short period of time.

[0027] In addition, the recorded changes in the R, G and B channels should take place at a sub-area, occupying a fraction of the total area of interest. For example, when the monitored area is a wide field of view to the horizon, an expected short-term spectral burst event (as resulting from a launch or an explosion) would not be reflected in all the pixels directed to the entire interest area, but rather localized to the sub-area in which the event takes place.

[0028] The range of pixels involved may a be minimal number of pixels anywhere between single pixel event to more extended area event covering tens of pixels for a closer event. For a typical full HD image comprising of 1920x1080 pixels this may be even below 0.001% extent of total observed area (i.e. the full image) per lOpixel area event. This minimal information extreme case option stresses the need to provide a strong spectral marker for a very minimal statistical base of data hereby explaining the significance of the system-algorithmic solution uniqueness / innovation.

[0029] Specifically, in the case of attempting to detect an event of a launch, a blast, an explosion or the firing of a firearm, the red channel will likely exhibit significant dominance over the green and blue channels. In addition, this change is likely to occur during a short period of time, manifesting as a sharp spike in the red channel (rapid rise and fall) compared to the green and the blue channels. These parameters, change, time and location, may be significant in distinguishing between a short-term spectral burst event and a simple sunset or bonfire. In the example of a moving and / or rotating arrangement of one or more image capturing devices as described above, the filtering of the set of pixels should account for the movement / rotation. In other words, in a moving / rotating configuration, a given pixel of the image capturing device is no longer directed to a constant point of the observed area.

[0030] Thus, the processing module may also include a registration module comprising an algorithm configured for compensating for the movement / rotation of the image capturing device, such that the filtered set of pixels is obtained for the correct points of the observed area. Thus, instead of obtaining data over time from a given pixel of the image capturing device, what is obtained is the data over time from different pixels all pointed at the same point in the observed area. The algorithm may also be configured for performing registration of a fixed coordinate system and associating each point in that coordinate system with changing pixels, monitoring them across one or more image capturing devices.

[0031] The one or more image capturing devices may be, for example, a still image camera taking series of still images, or a video device configured for outputting a set of image frames in a video format (e.g. a video file). In the latter case, it is common for video devices to compress videos based on various compression techniques (e.g. codecs).

[0032] In accordance with one example of the present application, the processing module may be configured for relying on the compression technique in order to obtain the filtered set of image frames. Specifically, instead of extracting the entire video (as is done automatically by any device running the video file), i.e. the entire raw data, and then filtering out the desired pixels for analysis therefrom, the processing module may rely on the known compression technique in order to directly extract only those pixels which may be relevant for detecting a burst-event. This, again, provides the advantage of drastically reducing the amount of data to be handled and analyzed, saving time, computational power and resources, and consequently costs.

[0033] In accordance with a specific example, the video file may be compressed in MPEG format, whereby the compression algorithm only records the changes in pixels between each two consecutive frames. In other words, it relies on the pixel variance - the change between pixels between consecutive frames. In general, MPEG is usually the video file container format that includes the compressed information. Nominal compression standards include h.264 and h.265 are the most popular.

[0034] Compression is usually comprised of data that compresses static reference image scene (usually called I frame for intraframe), that is compressed as is without any reference to previous or future frames very much like a JPEG compression. The remainder of the frames - actually most of the frames - are predicted based on that reference frame, and thereby allow to store only changes relative to reference rather than full pixel data thus minimizing stored / transmitted data rate

[0035] In accordance with the present application, an algorithm of the processing module may rely on this compression method, thereby extracting only varying pixels, similar to the filtered set of pixels previously discussed with respect to still images. Similar methods may be applied for other compression techniques.

[0036] The image capturing device may be a standard RGB camera, a video camera, and any other common image capturing device available on the market. It should be noted that the system does not require the use of a specially designed image capturing device in order to properly detect the desired short-term spectral burst-event. However, it should be appreciated that specifically designed cameras may be provided, allowing for increased image capture and compatibility with the system and method of detection.

[0037] In general, most RGB cameras comprise a silicon-based CCD or CMOS sensor that in sensitive to 0.2um to lum wavelengths. In order to extract only 0.4um-0.7um range that is relevant to human eye color physiological response, a special low pass IR cut filter is used to filter out NIR wavelengths between 0.7 to lum and also lower wavelengths relevant to UV radiation below 0.4um. This is done plus implementing alternating R / G / B color filters according to some bayer-based mask pattern scheme leads to RGB images usually transmitted as video or stills data stream.

[0038] In accordance with one variation of the present application, the RGB image capturing device may be configured to provide an NIR / IR performance boost by virtually removing this NIR / IR cut filter. This results in a unique sensor configuration allows boosting detection probabilities (for both NIR / IR), and at the same time allows better discrimination against non-valid targets as they will mostly radiate blue and green readings that are much higher or more similar to Red+NIR data response. In essence, this configuration allows leveraging the spectral sensitivity of the silicon detector as follows: First, removing the internal NIR blocking filter included in all VIS camera allows receiving (back) full VIS+NIR responsivity from 0.4um to lum. Then, a high pass filter may be added externally to receive an NIR only imaging device (e.g. adding NIR high pass filter after removing the NIR blocking filter). In this manner, a full NIR spectrum is provided from a cut on wavelength (e.g. 0.7um to lum or 0.76um to lum etc., depending on the cut on wavelength used for external add on filter).

[0039] It should be noted that more complex filters may be used, for example, notch filters that may avoid some special NIR wavelengths that can also be used for specific target attenuation of making them more discernable in data processing later on.

[0040] As the full spectrum of silicon is now responsive to the different RGB bayer pixels, the present configuration allows getting three new reading of the scene - B+NIR.B , G+NIR.G and R+NIR.R signals. The NIR.R , NIR.G and NIR.B signal parts stand for the residual NIR response of the corresponding pixels now that the IR CUT filter is removed. As the NIR.R is again much more strongly responsive relative to other NIR.B and NIR.G parts, it allows obtaining a stronger R vs G&B signals differing. Such a differentiation may be much more potent as it allows emphasizing the Red+NIR.R uniquely strong spectral signature and further differentiate it from the Blue+NIR.B and Green+NIR.G signals that are much less expected to compete with the Red+NIR boosted signal.

[0041] Last issue is the management of R / G / B signals not being co-located as they are placed one beside the other in the RGB bayer pixel filtering mask scheme. So per smaller target pixels, some kind of optical blurring effect may be added so the RGB reading is effectively more collocated in the signal reading thereby making the differentiation scheme better applied and more valid in terms of PD vs FAR performance needs.

[0042] It is also appreciated that converting an image device to an NIR receiver device allows for easier rejection of many sun induced signals by various method:

[0043] - testing for intensity levels (e.g. by canceling saturated signals that indicated direct solar reflection in NIR that cannot represent true fire event in relevant ranges that are expected by our system); analyzing color ratio or RGB / XYZ color plane locus; and raw RGB color plane after sRGB (standard RGB) plane data is converted back to raw RGB plane sensing by the silicon detector. The above method may allow achieving -1 : 100 FAR data reduction during sunlight intense reflection times, thereby making the detection system practical for use in true battlefield scenarios.

[0044] It should of course be noted that when changing the spectral filter from NIR blocking to NIR only sensing as mentioned above, the RGB color pixels in the sensor no longer signify VIS colors. Rather, they now relate to sensor bayer mask pixel filters residual spectral response in the NIR regime. This allows reorienting the RGB color pixels data to relate to discriminating feature of the fire vs. sun reflection discrimination.

[0045] For example, the B pixels response may represent deep NIR response (0.78um to lum), while the G and R pixels may represent more continuous response for full NIR spectra with a more pronounced response in different parts of NIR wavelengths.

[0046] In accordance with one example of the subject matter of the present application, the data obtained from the imaging device standard output may be internally processed sRGB color data space (as opposed to raw RGB), which may be used for discrimination context by specific work point calibration.

[0047] In accordance with another aspect of the subject matter of the present application there is provided a method for detecting a short-term spectral burst event using the system previously described, the method comprising the steps of: a. Receiving a filtered pixel set from an RGB sensor comprising R, G and B channels; b. Analyzing the R, G and B channels of at least a portion of said filtered pixel set; and c. Based on the spectral ratio between an intensity of the R channel and an intensity of the G and B channels, determining the existence of said short-term spectral burst transient event.

[0048] The filtered pixel set may be obtained by the steps of: d. Obtaining a set of image frames of a detection environment using an RGB spectral sensor, comprising R, G and B channels; and e. Filtering said set of image frames by removing non-varying pixels between each two consecutive image frames of at least a portion of said set of image frames, thereby generating a first filtered pixel set.

[0049] Alternatively, the filtered pixel set may be obtained by the steps of: f. Obtaining a video sequence of a detection environment using an RGB spectral sensor, comprising R, G and B channels; and g. By analyzing an encoding algorithm of said video sequence, filtering out non-varying pixels between each two consecutive images of at least a portion of said video sequence, thereby generating a second filtered pixel set.

[0050] The above method may also includes calibrating the data based on white balance, in particular, the camera provided sRGB data may be fixed for a white balance work point to optimize data discrimination performance and for making sure that color data produced by the camera is calibrated. Thereafter, the data may be changed to linear dynamic range data (i.e. reversing gamma correction), allowing to compute color change data in time (corresponding to change event detection). Finally, the simple color ratios may be used to discriminate true fire detection events from false events.

[0051] More intricate and optimal ways include converting sRGB color data to other color spaces and using color space segmenting or Al training methods to improve anomaly detection for any specific scenery or Combat zone performance.

[0052] It should be appreciated that a similar method to the above defined workflow may be used with RGB VIS sensor data as is, for example, by adapting it to fire / sun discrimination problem, by implementing proper data processing calibration.

[0053] In accordance with yet another aspect of the subject matter of the present application, there is provided a method for the detection of a spectral event, said method comprising the steps of: a. Obtaining a video sequence of a detection environment using an RGB spectral sensor, comprising R, G and B channels; b. By analyzing an encoding algorithm of said video sequence, filtering out non-varying pixels between each two consecutive images of at least a portion of said video sequence, thereby generating a filtered pixel set; c. Comparing the R, G and B channels of at least a portion of said filtered pixel set; and d. Based on the ratio between an intensity of the R channel and an intensity of the G and B channels, determining the existence of said spectral event.

[0054] It should be understood that the above referenced method may be used for the detection of various spectral events, either long-term or short-term, and relies on the significant reduction in the amount of data to be analyzed which is based on the compression technique rather than on the entire raw data of the video.

[0055] BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to better understand the subject matter that is disclosed herein and to exemplify how it may be carried out in practice, embodiments will now be described, by way of non-limiting example only, with reference to the accompanying drawings, in which:

[0057] Fig- 1 is a schematic illustration of a system in accordance with the present application;

[0058] Fig- 2 is a schematic illustration of an image frame recording the area observed by the system shown in Fig. 1, divided into areas;

[0059] Fig- 3 is a schematic illustration of a spectral event occurring in the area observed by the system of Fig. 1;

[0060] Figs. 4A-4E are schematic illustrations of a sequence of image frames obtained by the system, recording the spectral burst event shown in Fig. 3;

[0061] Figs. 5A to 5E; are schematic illustrations of a filtered set of pixels taken from the image frames shown in Figs. 4A-4E;

[0062] Fig. 6A is a schematic illustration of a first RGB analysis of the filtered pixel set shown in Figs. 5A-5E over time;

[0063] Fig. 6B is a schematic illustration of a second RGB analysis of the filtered pixel set shown in Figs. 5A-5E over time;

[0064] Fig. 7 is a schematic block diagram of the method of the present application reliant on still images;

[0065] Fig. 7B is a schematic is a schematic block diagram of the method of the present application reliant on a video file;

[0066] Fig. 8A is a schematic top view illustration of another example of the system of the present application;

[0067] Fig. 8B is a schematic illustration of another example of the system of the present application; and

[0068] Figs. 9A-9C is a schematic illustration of a set of image frames obtained by the system shown in Fig. 8B. It will be appreciated that for simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn accurately or to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements for clarity, or several physical components may be included in one functional block or element. Further, where considered appropriate, reference numerals may be repeated among the figures to indicate corresponding or analogous elements.

[0069] DETAILED DESCRIPTION OF EMBODIMENTS

[0070] Attention is first drawn to Fig. 1 in which a system for the detection of spectral events is shown, generally designated 1, and comprising an image capturing device 10, and a processing module 100. The system 1 is shown with the image capturing device 10, directed towards a scene in which such a spectral event is to be detected.

[0071] The image capturing device 10 has a given field of view which allows it to capture images of a portion of the scene, referred to herein as the observed area 30 of image capturing device 10. The image capturing device 10 may be a standard camera, outputting image frames in the form of raw RGB data.

[0072] With additional reference being made to Fig. 2, the observed area 30 is divided into clusters 40 of pixels 50, in the form of a rectangular grid 60. Each cluster 40 is shown as a square on the grid 60, and contains therein a plurality of pixels 50 (shown Figs. 5A to 5E).

[0073] The images acquired by the image capturing device 10 are provided to the processing module 100 for analysis.

[0074] During a static state, when no spectral event occurs in the scene, within the observed area 30, no change will be recorded in the pixels 50 of the clusters 40 between successive image frames provided to the processing module. However, upon the occurrence of a spectral event S, this event will be recorded as a change in the properties of the pixels 50 (e.g. brightness, intensity, hue, contrast, time, and location), as shown on cluster 40;,7of the grid 60 in Fig. 3.

[0075] Attention is now drawn to Figs. 4A to 4E, showing the observed area at five different instances in time during the occurrence of a spectral event S. As observed, Fig. 4A captured the state of the scene before the occurrence of the spectral even S. Figs. 4B to 4D capture the spectral event S as it expands, during four different instances Sti, St2, St3 and St4, covering more of the area of the cluster 40y, and Fig. 4E captures the spectral event during its dying stages, when its area shrinks to cover less of the cluster 40ij.

[0076] The processing module 100 first processes the image frames obtained from the image capturing device 10, and filters out (i.e. ignores) any clusters 40 in which no change has been recorded between subsequent images, and, moreover, any pixels 50 within the cluster 40 in which no change has been recorded, as shown in Figs. 5A to 5E.

[0077] Figs. 5A to 5E, correspond to Figs. 4A to 4E, but show only the cluster 40ij, since this is the only cluster of the observed area 30 in which a change has been recorded. In addition, within the cluster 40ij, the varying pixels 52 are shown in grey, while the non-varying pixels 54 are shown in white.

[0078] Extracting the filtered set of varying pixels 52 drastically reduces the number of pixels 50 to be analyzed by the processing module 100, which now only has to process a tiny fraction of the entire data of the images in order to determine the occurrence of the spectral even S. Specifically, the processing module 100 of the present example now needs to deal with only one cluster out of the original hundred and twenty clusters 40, and even less so, only with a select number of pixels (varying pixels 52) of that single cluster 40ij.

[0079] This provides the advantage of reduced computing power, which in turn, results in reduced latency, higher precision, cost effectiveness etc., to name a few.

[0080] Turning now to Figs. 6A and 6B, the processing module 100 is now tasked with determining whether the changes recorded in the filtered pixel set 52 is indicative of an occurrence of an anticipated spectral even S.

[0081] In this example, the system 1 is shown configured for the detection of a shortterm spectral burst event SB, which may be the result of a blast, a launch of a rocket, firing from a firearm, etc. Specifically, such events SB are characterized by occurring over an extremely short period of time (seconds or fractions thereof), and exhibiting a sharp rise and fall in terms of spectral characteristics.

[0082] Observing Fig. 6A, an RGB analysis is shown plotted on a graph in which the X-axis denotes time (t) and the Y-axis denotes the intensity I. The analysis of the processing module 100 detects a significant change in the red channel R compared to the green and blue channels G, and which is more, this change is in the form of a peek P occurring over a short period of time (e.g. 0.4 seconds). Specifically, the R channel exhibits an intensity which is above an expected threshold T, the ratio between the intensity of the R channels and the intensities of the G and B channels is significant, and the sharp rise and fall may be indicative of the occurrence of a short-term spectral burst event SB. The analysis thus indicates that likelihood of such an event is high.

[0083] As shown in Fig. 6B, depicting another example of an event, the intensity of the red channel R similarly crosses the expected threshold T. However, the ratio between the intensity of the R channels and the intensities of the G and B channels is less significant, and the entire event takes place over an extended period of time (10 seconds), thereby indicating that the observed spectral event may not be a short-term spectral burst event SB, but rather another type of event (e.g. fireworks).

[0084] In processing the data from the image capturing device 10, the processing module may take into account a variety of parameters, which may include, but are not limited to: intensity, hue, brightness, contrast, position (which pixels are varying), time, variance pattern etc.

[0085] With reference being made to Fig. 7A, a block diagram of the above discussed detection method is shown. Specifically, the image capturing device 10 acquires a set of image frames and provides them to the processing module 100. The filtering module 120 of the processing module 100 then filters the set of image frames and the analysis module 140 performs the analysis to determine the occurrence of a spectral event.

[0086] However, as shown in Fig. 7A, the processing module 100 may also rely on other methods for obtaining the filtered pixel set. Specifically, the system 1 may comprise a video image capture device 10’, also RGB based, which may be configured for outputting a video file under a given video encoding. A video encoding allows compressing the raw data acquired by the video image capturing device 10’ in order to reduce the amount of data to be transmitted, resulting in a smaller size of digital files.

[0087] With additional reference being made to Fig. 7B, one example of such encoding is MPEG, in which the raw video data is encoded (12’) to only record changes to each pixel, i.e. pixel variance, rather than providing the entire pixel data of each frame of the video. This results in a considerably smaller encoded video file 14’. Under normal operating circumstances, a compatible computer software is configured for receiving the encoded MPEG video file, decode it and extract the raw data.

[0088] However, under the method of the present invention, instead of decoding the data, a filtering module 120’ of the processing module 100 is configured to only examine the encoded MPEG video file, and extracting therefrom, based on the encoding, only the changes in the pixels. In others words, instead of extracting the raw data and performing an analysis thereon, the processing module 100 is configured for directly extracting only the varying pixels 52, thereby easily creating the filtered pixel set 52’.

[0089] The second filtered pixel set 52’ can then be analyzed by the processing module 100 in a similar manner to that performed on the first pixel set 52 obtained from the still image capturing device 100, e.g. a still image camera.

[0090] This method has the similar advantage of reducing the required computing power, reducing latency etc.. It should also be noted that the method is not limited to an MPEG encoding and may be adapted to various encoding methods, thereby eliminating the need to extract all the raw data before performing the analysis.

[0091] Attention is now drawn to Fig. 8A, in which another example of a system is shown, generally designated 1”, and comprising a plurality (four) image capturing devices 10, arranged in order to cover 360° of the observed scene.

[0092] Under this example, the processing module may consolidate the data received from all image capturing devices 10, or, alternatively, the system 1” may comprise an individual processing module 100 for each image capturing device 10.

[0093] Turning now to Fig. 8B, another example of the system is shown, generally designated 1’”, in which a single image capturing device 10 is shown mounted on a gimbal 12, allowing it to rotate 360° to cover all of the observed scene.

[0094] Under this example, following the changes in a given cluster 40 or pixel 50 of the image capturing device 10 would not suffice, as the same cluster / pixel constantly change their position and the area of the scene which they record.

[0095] Thus, with additional reference being made to Figs. 9A to 9D, the processing module 100 may comprise a coordinate registration associated with the gimbal 12, and allowing the processing module 100 to always refer to the relevant cluster 40 directed at a constant point of the scene.

[0096] Specifically, in the case of a spectral event, the processing module 100 knows that the given point in the scene is represented by cluster 40ij in Fig. 9B, cluster 40i,h in Fig. 9C and cluster 40i,f in Fig. 9D. As a result, when extracting the filtered pixel set 52 from the image frames, the processing module 100 will focus on the changes over time between different clusters 40ij to determine the occurrence of a spectral burst event. Those skilled in the art to which this invention pertains will readily appreciate that numerous changes, variations, and modifications can be made without departing from the scope of the invention, mutatis mutandis.

Claims

CLAIMS:

1. A system for the detection of a short-term spectral burst event, said system comprising a processing module configured for: receiving raw RGB data in the form of a set of image frames obtained by an image capturing device which comprises at least one sensor comprising R, G, and B channels; filtering said set of image frames by removing non-varying pixels between each two consecutive image frames of at least a portion of said set of image frames, thereby extracting a first filtered pixel set comprising a reduced amount of data compared to said set of image frames; performing a spectral analysis of the R, G and B channels of at least a portion of said first filtered pixel set based on a spectral ratio between properties of the R channel and properties of the G and B channels, over time; and based on said spectral analysis, determining the existence of said short-term spectral burst event.

2. A system according to Claim 1, wherein the system is configured for detecting at least one of: a 'short-term spectral burst event' and a 'long-term spectral event'.

3. A system according to Claim 2, wherein said short-term spectral burst event lasts between 1msec or sub 1msec to 60msec.

4. A system according to Claim 2, wherein said short-term spectral burst event is a result of at least one of the following: a burst of fire, launch of a missile or rocket, fire from a firearm, a blast, and an explosion.

5. A system according to Claim 2, wherein the spectral event takes up between 0.001% to 2% of the total observed area, more particularly 0.01% to 1.8%, and even more particularly between 0.1% to 1.2% of the total observed area.

6. A system according to Claim 2, wherein said long-term event is a result of at least one of the following: a forest fire, a sunset, or fireworks.

7. A system according to any one of Claims 1 to 6, wherein the system further comprises an image capturing device with at least one sensor comprising R, G, and B channels, and is configured for obtaining said raw RGB data and outputting it to said processing module.

8. A system according to Claim 7, wherein the system comprises an image capturing arrangement comprising one or more such image capturing devices, configured for coverage of a wide area of interest.

9. A system according to Claim 8, wherein said arrangement is such that combined areas of viewing of each image capturing device are sufficient for covering a desired area of interest of the entire system.

10. A system according to Claim 8, wherein the system is configured for moving and / or rotating, such that each of the one or more image capturing devices browses over an entire area of interest of the system, each time capturing images of a different portion of the entire area of interest.

11. A system according to Claim 1, wherein the processing module comprises a filtering module configured for extracting a filtered pixel set from the set of image frames.

12. A system according to Claim 11, wherein the processing module comprises an analysis module configured for analyzing the filtered pixel set extracted by the filtering module.

13. A system according to Claim 1, wherein the spectral analysis is based on any one or a combination of at least the following parameters: brightness, intensity, hue, contrast, time, and location.

14. A system according to Claim 1, wherein the RGB data is compared with a threshold.

15. A system according to Claim 10, wherein the processing module includes a registration module comprising an algorithm configured for compensating for movement / rotation of the image capturing device.

16. A system according to Claim 1, wherein the one or more image capturing devices are at least one of a still image camera, and a video device.

17. A system according to Claim 1, wherein the processing module is configured for relying on a compression technique of a video device in order to directly extract the filtered pixel set.

18. A system according to Claim 17, wherein said encoding is an MPEG encoding.

19. A system according to Claim 1, wherein the image capturing device comprises a virtually removed NIR / IR cut filter.

20. A system according to Claim 1, wherein the system comprises a virtually added high-pass filter for removing an RGB VIS range from the sensor.

21. A method for detecting a spectral event using a detection system, the method comprising the steps of a. receiving a filtered pixel set from an RGB sensor comprising R, G and B channels; b. analyzing the R, G and B channels of at least a portion of said filtered pixel set; and c. based on a spectral ratio between an intensity of the R channel and an intensity of the G and B channels, determining the existence of said shortterm spectral burst transient event.

22. A method according to Claim 21, wherein the filtered pixel set is obtained by the steps of d. obtaining a set of image frames of a detection environment using an RGB spectral sensor, comprising R, G and B channels; and e. filtering said set of image frames by removing non-varying pixels between each two consecutive image frames of at least a portion of said set of image frames, thereby generating a first filtered pixel set.

23. A method according to Claim 21, wherein the filtered pixel set is obtained by the steps of f. obtaining a video sequence of a detection environment using an RGB spectral sensor, comprising R, G and B channels; and g. by analyzing an encoding algorithm of said video sequence, filtering out non-varying pixels between each two consecutive images of at least a portion of said video sequence, thereby generating a second filtered pixel set.

24. A method according to Claim 23, wherein: raw data of said video sequence comprises a set of raw image frames; said video sequence is encoded with an encoding algorithm based on a pixel variance between consecutive raw image frames; and said second filtered pixel set is based on said pixel variance.

25. A method according to any one of Claims 20 to 24, wherein said spectral event is depicted by pixels in an event area of an image frame, and occupies no more that 2% of the total observed area of said image frame.

26. A method according to any one of Claims 20 to 24, wherein said analyzing comprises detection of an increase and a decrease in said spectral ratio over time for detection of a short-term spectral burst within an event area.

27. A method for the detection of a short-term spectral burst event, said method comprising the steps of: a. obtaining a set of image frames of a detection environment using an RGB spectral sensor, comprising R, G and B channels; b. filtering said set of image frames by removing non-varying pixels between each two consecutive image frames of at least a portion of said set of image frames, thereby generating a filtered pixel set; c. comparing the R, G and B channels of at least a portion of said filtered pixel set; and d. based on the ratio between an intensity of the R channel and an intensity of the G and B channels, determining the existence of said short-term spectral burst event.

28. A method for the detection of a short-term spectral event, said method comprising the steps of: e. obtaining a video sequence of a detection environment using an RGB spectral sensor, comprising R, G and B channels; f. by analyzing an encoding algorithm of said video sequence, filtering out non-varying pixels between each two consecutive images of at least a portion of said video sequence, thereby generating a filtered pixel set; g. comparing the R, G and B channels of at least a portion of said filtered pixel set; and h. based on the ratio between an intensity of the R channel and an intensity of the G and B channels, determining the existence of said spectral event.

29. A non-transitory computer readable storage medium having computer readable program code embodied therewith, the computer readable program code, executable by at least one processing circuitry of a computer, the method comprising: a. receiving a filtered pixel set from an RGB sensor comprising R, G and B channels; b. analyzing the R, G and B channels of at least a portion of said filtered pixel set; andc. based on a spectral ratio between an intensity of the R channel and an intensity of the G and B channels, determining the existence of said shortterm spectral burst transient event.

30. A method of converting an imaging device into a detection system for the detection of a short-term spectral burst event, said method comprising the steps of: a. obtaining an imaging device comprising an RGB sensor sensitive to the VIS range; b. modifying the sensor to become at least partially sensitive to the NIR / IR range; and c. modifying the sensor to at least partially remove sensor sensitivity to RGB in the VIS range.

31. A method according to Claim 30, wherein step (b) is performed by removing a built-in NIR / IR cut-off filter provided with the sensor.

32. A method according to Claim 30, wherein step (c) is performed by applying a high-pass filter.

33. A method according to Claim 30, 31 or 32, wherein the resulting detection system is sensitive to the NIR / IR range only.

Images