Method for Tracking the Location of a Radiation Source

Abstract

A method for generating a reconstructed image that shows the position of a radiation source in an environment, the reconstructed image being based on a gamma image acquired by a gamma camera that is sensitive to ionizing electromagnetic radiation and that can be moved relative to at least one radiation source between two different measurements, the gamma camera being connected to a visible light camera configured to form a visible light image of the environment, the gamma camera and the visible light camera defining a field of view, the method including generating the reconstructed image showing the position of at least one radiation source in the field of view, the gamma camera and the visible light camera being moved between at least two measurements.

Description

Technical Field

[0001] The technical field of the present invention is the characterization of radiation sources present in the environment, particularly in nuclear facilities or facilities including radiation sources. Background Art

[0002] Gamma cameras are devices that allow the mapping of radiation sources in a given environment, particularly in nuclear facilities, using images. Such devices were developed in the 1990s and are increasingly used in nuclear facilities for radiological characterization purposes. The aim is to identify the main radiation sources present in the facility. Specifically, the radiation sources are not evenly distributed. They are typically concentrated locally, forming "hot spots", using the terminology commonly used in the field of radiation protection. The advantage of gamma cameras is that they allow the localization of these hot spots at a certain distance.

[0003] The improvement and use of gamma cameras have been widely described in the literature. Since the early 2000s, spectroscopic gamma cameras have been under improvement. These cameras are based on pixelated imagers, and each pixel allows the acquisition of a spectrum from the radiation it detects. Thus, it is easier to localize radiation sources. Specifically, the spectroscopic function allows the selection of the energy band of interest, which corresponds to non-scattered photons, i.e., photons that have not deviated since they were emitted by the radiation source. The path of non-scattered photons is straight. Selecting them in a predetermined energy band allows the removal of the noise corresponding to scattered photons. Since the latter photons have deviated since their emission, they do not provide useful information about the position of the radiation source. Thus, scattering is a noise source that can be significantly limited by spectroscopy.

[0004] Another advantage of spectroscopic gamma cameras is that knowledge of the photon energy allows the identification of the isotopes responsible for the radiation. This is very important information in the field of radiation protection, or in radioactive waste management, or even in radiological characterization during the dismantling of nuclear facilities or after an accident.

[0005] One limitation associated with the use of gamma cameras is that it is not possible to focus X-rays or gamma rays with energies exceeding a few keV or even a few hundred keV. Thus, the collection efficiency of X-ray or gamma photons is low. The collection efficiency corresponds to the number of photons reaching the detector, which is normalized by the number of photons emitted. Another difficulty is the low detection efficiency of detectors sensitive at such energies, especially when compact detectors are preferred. The combination of low collection efficiency and low detection efficiency results in the need to usually acquire images with a relatively long acquisition time in order to increase the number of detected photons. Thus, gamma cameras are usually kept stationary for several seconds or even tens of seconds in order to obtain sufficient measurement statistics.

[0006] Certain improvements have been made to allow the generation of a three-dimensional reconstruction of the position of a radiation source. This is a matter of applying the principle of triangulation while moving a gamma camera between the acquisition of various images and taking multiple images of the same radiation source. However, this method requires high computational power. In addition, this method does not allow the movement of the radiation source in a stationary environment to be taken into account.

[0007] The inventors have proposed a method that is simple and inexpensive in terms of memory size and computational power and allows the movement of the gamma camera relative to a stationary radiation source, or the movement of the radiation source relative to the gamma camera, to be taken into account. This method allows satisfactory gamma images to be acquired while moving the gamma camera relative to one radiation source or relative to multiple radiation sources. Summary of the Invention

[0008] A first subject of the present invention is a method for forming a reconstructed image that shows the position of a radiation source in a field of view, the method using a device that includes a gamma camera connected to a visible light camera, the device being such that:

[0009] - The gamma camera and the visible light camera define the field of view;

[0010] - The visible light camera is configured to form a visible light image of the field of view;

[0011] - The gamma camera includes pixels that are configured to detect ionizing electromagnetic radiation generated by a radiation source potentially present in the field of view, the pixels being located in a detection plane;

[0012] - The gamma camera is configured to form a gamma image that allows the position of the radiation source in the field of view to be estimated, the radiation of the radiation source being detected by the pixels;

[0013] The method includes the following iterative steps:

[0014] a) At an initial measurement time, an initial gamma image is acquired with the gamma camera and an initial visible light image is acquired with the visible light camera;

[0015] b) Based on the initial gamma image and the initial visible light image, the reconstructed image is initialized;

[0016] c) At a measurement time, a gamma image is acquired with the gamma camera and a visible light image is acquired with the visible light camera;

[0017] d) The visible light image at this measurement time is compared with the visible light image at a previous time, the previous time being the initial measurement time or the previous measurement time, and then, based on the comparison, the motion field between the visible light images acquired at the previous time and the measurement time is estimated;

[0018] e) The reconstructed image is updated based on the following:

[0019] - a reconstructed image established at a previous time;

[0020] - a motion field obtained from step d);

[0021] and a gamma image acquired at the measurement time;

[0022] f) Optionally, superimpose the reconstructed image or a part of the reconstructed image on a visible light image;

[0023] g) While increasing the measurement time, repeat steps c) to f) until the iteration is stopped;

[0024] Between at least two measurement times, move the device such that one field of view corresponds to each measurement time.

[0025] Thus, the reconstructed image shows the position of the radiation source in the field of view. According to one embodiment, the reconstructed image at the measurement time is defined only in the field of view corresponding to the measurement time.

[0026] Preferably, between at least two measurement times, move the device such that the fields of view before and after the movement respectively include a common part.

[0027] The gamma camera defines a gamma field of view. The visible light camera defines a visible light field of view. The field of view of the device corresponds to the intersection of the gamma field of view and the visible light field of view.

[0028] The gamma image allows the localization of gamma sources potentially present in the gamma field of view.

[0029] According to a preferred embodiment, at each measurement time, each pixel that has detected the radiation generated by the radiation source is associated with a position in the object surface disposed facing the detection plane, and the reconstructed image corresponds to the spatial distribution of the radiation source on the object surface within the field of view at the measurement time. The object surface may advantageously be an object plane disposed parallel to the detection plane, and each position is coplanar.

[0030] According to this embodiment, the reconstructed image at the measurement time is defined only in the intersection of the field of view at the measurement time and the object surface. The reconstructed image at the measurement time may be defined only in the intersection of the field of view at the measurement time and the object surface.

[0031] Step e) may include:

[0032] i) Estimate the reconstructed image at the measurement time based on the reconstructed image established at the previous time and the motion field obtained in step d);

[0033] ii) Estimate the gamma image at the measurement time based on the estimated reconstructed image;

[0034] iii) Compare the gamma image estimated at the measurement time with the gamma image acquired at that measurement time;

[0035] iv) Update the reconstructed image at the measurement time based on the comparison.

[0036] The method may include determining a registration function representing the spatial offset between the gamma camera and the visible light camera, which is used in step e). The registration function allows the motion field of the gamma image to be defined based on the motion field obtained in step d). This allows the motion of the two gamma images acquired at said times to be obtained based on the motion established for two visible light images at a previous time and at the measurement time.

[0037] Step d) may include selecting notable points in the visible light images acquired at the previous time and at the measurement time, and the motion field includes the two-dimensional motion of the selected notable points.

[0038] The notable points may preferably be considered to belong to the same surface, or even to the same plane. This may particularly be a matter of the object surface or the object plane or a plane parallel to the object plane.

[0039] The method may be such that step d) at each measurement time includes:

[0040] - Extract a first notable point from the visible light image acquired at the measurement time;

[0041] - Extract a second notable point from the visible light image acquired at the previous time;

[0042] - Match the first notable point with the second notable point to form paired matching points, each pair being formed by a first notable point and a second notable point;

[0043] - Determine a motion for each pair of matching points;

[0044] - Obtain the motion field based on the motion determined for each pair of matching points.

[0045] The method may include:

[0046] - Generate a first grid of the observation field at the measurement time using the first notable points, and the first notable points form the first vertices of the first grid;

[0047] - Generate a second grid of the observation field at the previous time using the second notable points, and the second notable points form the second vertices of the second grid;

[0048] - Determine the motion at points belonging to the first grid and the second grid by interpolation according to the motion determined for each pair of notable points;

[0049] - Determine a motion field based on the motion of each matched vertex and the motion of the insertion.

[0050] According to one embodiment:

[0051] - The device is coupled to a motion sensor to estimate the motion of the device between two consecutive times;

[0052] - Step d) takes into account the estimated motion to estimate or verify the motion field.

[0053] One aspect of the present invention is that the motion field established using visible light images is considered to represent the motion of the radiation source between two measurement times, and the registration function compensates for any differences.

[0054] According to one embodiment, the gamma camera is configured to simultaneously acquire gamma images in various energy bands in steps a) and c), and steps b) and d) to f) are respectively implemented for each energy band so as to obtain reconstructed images in various energy bands at each measurement time.

[0055] According to one embodiment, the gamma camera is configured to simultaneously acquire gamma images in various energy bands in steps a) and c), the method includes acquiring a linear combination of various gamma images in various energy bands at the same measurement time so as to obtain a combined gamma image, and performing steps b) and d) to f) based on the combined gamma image formed at each measurement time. The combined image may be formed by a weighted sum of gamma images acquired at various energies, and the weighted sum uses a weighting factor determined based on the emission spectrum of the radioisotope such that each reconstructed image represents the spatial distribution of the activity of the isotope in the observation field corresponding to the measurement time.

[0056] Another subject of the present invention is a device, comprising:

[0057] - A gamma camera, which includes pixels and defines a gamma observation field, and the gamma camera is configured to determine the position of the source of X-ray or gamma radiation in the gamma observation field;

[0058] - A visible light camera, which allows the formation of visible light images of the visible light observation field;

[0059] - The intersection of the gamma observation field and the visible light observation field is not zero, and defines the observation field of the device;

[0060] - An image processing unit, which is configured to receive at each measurement time:

[0061] · Gamma images, which are formed by the gamma camera;

[0062] · Visible light images, which are formed by the visible light camera;

[0063] The processing unit is configured to execute steps b) and d) to g) of the method according to the first subject matter of the present invention.

[0064] The present invention will be better understood by reading the description of the exemplary embodiments described in the remainder of the specification and by referring to the accompanying drawings listed below. Description of the Drawings

[0065] Figure 1A A measuring device arranged facing a structural element and a radiation source is schematically shown.

[0066] Figure 1B and 1C A gamma image and a visible light image acquired at a single measurement time are respectively shown.

[0067] Figure 2A A measuring device arranged facing a structural element and a radiation source at another measurement time is schematically shown.

[0068] Figure 2B and 2C A gamma image and a visible light image acquired at another measurement time are respectively shown.

[0069] Figure 3 A plane of an object relative to a gamma camera is schematically shown.

[0070] Figure 4 The main steps of a method for obtaining a reconstructed image are shown. Detailed Description of the Invention

[0071] Figure 1A A measuring device 1 allowing the implementation of the present invention is shown. The measuring device includes a gamma imager 2 or a gamma camera. The gamma imager is configured to detect ionizing electromagnetic radiation of the X-ray or gamma-ray type, the energy of which is typically from 10 keV to 10 MeV. The gamma imager includes pixels, each pixel corresponding to a basic spatial region of the field of view. The pixels are located in a detection plane P. When the basic spatial region corresponding to a pixel includes a source of X-ray or gamma radiation, some of the radiation emitted by the source reaches the pixel and is detected by the pixel. Thus, the amplitude of the signal of the pixel of the gamma image increases as the basic spatial region associated with the pixel becomes more radioactive (i.e., emits more X-ray or gamma radiation). In the remainder of the specification, examples related to gamma radiation sources are given, which correspond to the most common application scenarios. It can be directly converted to an X-ray radiation source.

[0072] The gamma imager can be a Compton gamma camera, a pinhole collimator gamma camera or a coded aperture gamma camera. This can also be a non-limiting matter of a gamma camera whose collimator includes parallel channels, or converging channels, or diverging channels. Thus, the term gamma camera corresponds to an imager having a field of view and configured to form an image allowing the localization of a radiation source in a radiation field. Whatever the type of gamma imager, it allows the formation of a gamma image including pixels, each pixel corresponding to an elementary spatial region of the observed environment. Certain gamma imagers have a spectrometric function, in the sense that they allow the spectral separation of the radiation detected in the various spectral bands. When using this type of imager, various gamma images of a given field of view can be formed, each image corresponding to a spectral band. Images can also be established by considering a combination of spectral bands corresponding to the emission spectrum of the isotope. This combination can be a weighted sum. Then, the image represents the spatial distribution of the activity of the relevant isotope.

[0073] In the case of certain gamma imagers, especially Compton gamma cameras or coded aperture gamma cameras, the images acquired by the imager do not allow direct observation of the emission sources in the field of view. Taking into account the response function of the camera, the acquired images are processed in order to allow the obtaining of a gamma image in which the brightness of each pixel corresponds to the emission intensity of the elementary spatial region of the field of view associated with the pixel.

[0074] Gamma cameras are conventionally associated with visible light cameras. These are standard cameras allowing the formation of a visible light image of the field of view. Figure 1A The device shown includes a visible light camera 3 firmly fixed to the gamma camera 2. Both the visible light camera 3 and the gamma camera 2 have a field of view extending around their respective optical axes. In Figure 1A the optical axes and the fields of view of the gamma camera 2 and the visible light camera 3 are represented as Ω2, Δ2, Ω3 and Δ3. Generally, the visible light field of view is larger than the gamma field of view. The visible light field of view and the gamma field of view overlap: their intersection is non-zero. The latter is as large as possible.

[0075] The visible light camera 3 is generally placed at a small distance from the gamma camera such that their respective optical axes are close together and preferably parallel to each other. Thus, beyond a certain distance, generally less than 1 m, or even less than 50 cm, the field of view of the gamma camera is included in the field of view of the visible light camera. The intersection of the two fields of view forms the field of view Ω of the device.

[0076] Preferably, the optical axes of the gamma camera 2 and the visible light camera 3 are aligned. The gamma camera and the visible light camera are calibrated to correct the parallax error due to the offset of the two optical axes, and to take into account the geometric distortion of the lens of the visible light camera, especially at the edges of the field of view. The calibration also takes into account the differences in the number and size of pixels between the visible light camera and the gamma camera. This calibration allows the registration function to be defined.

[0077] The processing unit 4 receives the images generated by the gamma camera 2 and the visible light camera 3. The image processing unit is specifically configured to merge the image obtained from the visible light camera 3 and the image obtained from the gamma camera 2, particularly by taking into account the registration function. The aim is to obtain a synthetic image corresponding to the visible light image, but on which the radiation sources detected by the gamma camera appear. The processing unit 4 is connected to a memory in which the image processing instructions are encoded. The processing unit 4 includes a processor (e.g., a microprocessor) configured to execute the instructions corresponding to the steps described below.

[0078] As mentioned with respect to the prior art, the acquisition time of the gamma image is typically several seconds, or even dozens of seconds. This is due to the low collection efficiency and the low detection efficiency.

[0079] Figure 1A The radiation sources 10 are shown a 、10 b and 10 c , which are the emission sources of γ radiation, and the structural elements 11 a 、11 b and 11 c schematically represent the visible environment around the radiation sources. The device 1 occupies a position and has an orientation in the observed environment. Figure 1B Schematically shown is the gamma image obtained by the gamma camera. In the said image, by way of illustration, the radiation sources 10 present in the field of view are represented in the form of circles a and 10 b . Figure 1C Schematically shown is the visible light image obtained by the visible light camera, and it allows the observation of the objects present in the field of view. The term object corresponds to the structural elements present in the field of view. It can be a part of a building element (wall, door, window, ridge) and / or a part of an industrial installation (pipe, specific parts of equipment) and / or a furniture item (table, chair, box, various items).

[0080] Figure 2A is shown relative to Figure 1ADevices shown are in the same environment, in the same position and in different orientations. Generally, due to the sensitivity of the gamma camera, acquisitions lasting several seconds or even minutes are required to obtain a representation of each radiation source with sufficient signal-to-noise ratio. During the acquisition of a gamma image, the gamma camera must preferably remain stationary.

[0081] Figure 2B The acquired gamma image is shown. The circles drawn in solid lines show the radiation sources 10 a 、10 b and 10 c . In this image, the dashed lines show the positions of the radiation sources 10 Figure 2B as they appear in a and 10 b . In the gamma image, a change in the orientation of the gamma camera between Figure 2A and 2B results in the movement of the radiation sources, which is represented by two arrows. One can consider an environment in which each radiation source detected by the gamma camera has a fixed position, which is represented by two-dimensional coordinates in the object space. When the gamma camera moves in the object space, the gamma sources present in the field of view are detected and generate exploitable signals in the gamma image. Generally, the signals formed in the gamma image have a relatively low signal-to-noise ratio, which depends on the radiation level generated by the sources on the gamma camera. To improve the signal-to-noise ratio, the signals generated by the radiation sources on the gamma camera can be accumulated, for example, by moving average. Thus, each time a gamma source is present in the field of view, the signal it generates in the gamma image is taken into account in order to improve the detection statistics associated with this source. Then, gamma images acquired previously and / or subsequently are considered and integrated into a reconstruction method.

[0082] A visible light image acquired by a visible light camera is associated with each acquisition of a gamma image. The visible light camera 3 acquires a visible light image of the observed scene. The field of view Ω of the imaging device, i.e. at the intersection of the respective fields of view of the gamma camera and of the visible light camera, defines the object space. The object space contains points (x, y), each of which is associated with a pixel (u, v) of the gamma image. The pixels belong to the detection plane P in which the gamma image is formed. After taking into account the registration function, the points (x, y) of the field of view have two-dimensional coordinates and correspond to the pixels of the visible light image.

[0083] An important element of the invention is that the points of the object reference frame are considered to belong to the same object projection plane P o . The angular field of view Ω2 of the gamma camera extending around the optical axis Δ2 describes a part of the sphere S (see Figure 3 ). The object plane P ocorresponds to a plane that is tangent to the sphere S and perpendicular to the optical axis Δ2. The radiation sources present in the field of view are considered to be coplanar and belong to the object plane. The distance between the detection plane and the object plane is an arbitrary distance, which may be unknown. This distance is not taken into account in the reconstruction method described below. According to one variant, the radiation sources are considered to belong to the same three-dimensional surface arranged facing the detection plane P.

[0084] After image processing, it is allowed to perform image merging (thresholding and / or superimposition, where one image is visible through the other) and to consider the registration function, and the radiation levels associated with the points in the object reference system and considered to be significant are superimposed in the form of color codes on the visible light image V k on.

[0085] Thus, the field of view Ω of the device delimits a part of the object plane in which the reconstruction is performed.

[0086] Object plane P O The correspondence between each point (x, y) of the object plane P and each pixel (u, v) of the gamma image depends on the spatial response function of the gamma camera. When gamma photons are emitted from a point (x, y) in the object reference system towards the gamma camera, the trajectories they form in the gamma image are generated by the spatial response function F of the camera. When using a gamma camera based on a pinhole collimator, the spatial response function takes into account the aperture of the pinhole. It can be approximated by the image of a radiation source at the center of the field of view, which is used in the form of an impulse response. When using a coded aperture gamma camera, the response function takes into account the geometry of the mask. When using a Compton gamma camera, the response function depends on the detected energy and the position of the pixel in the detection plane P that has detected the radiation. Thus, in a Compton gamma camera, the response function can vary at each measurement time.

[0087] The correspondence between the points in the field of view and the pixels of the gamma image can be determined, for example, by convolution of the gamma image with the response function (especially in the case of a gamma camera equipped with a pinhole collimator) or by linear projection (especially in the case of a gamma camera equipped with a coded aperture collimator). The projection can also be of another type, such as but not limited to spherical projection or orthographic projection.

[0088] Thus, the response function F makes it possible to go from the detection plane in which the gamma image is acquired to the object plane P corresponding to the observed scene and in which the position of the radiation source is sought o . In the rest of this specification, the symbol F corresponds to the projection of the image formed in the object plane towards the detection plane. The symbol F — corresponds to the inverse projection of the image formed in the detection plane P towards the object plane P o . In the first method, F is a linear operator, and F— Corresponding to the transpose of F.

[0089] If M k is the gamma image acquired at time t k and I k is the reconstructed image showing the radiation sources in the field Ω k observed at time t k in the object plane, the following formula is obtained:

[0090] I k (x, y) = F - [M k (u, v)] (1)

[0091] Where:

[0092] (x,y) are the coordinates in the object plane parallel to the detection plane and corresponding to the coordinates (u,v) in the detection plane;

[0093] I k (x,y) corresponds to the image of the distribution of the radiation sources in the observation field. It is a two-dimensional matrix problem, and each item of the two-dimensional matrix represents the emissivity of the radiation source in the object plane at the measurement time t k at the coordinates (x,y) in the spectral band or multiple combined spectral bands. As described above, when there are multiple spectral bands in the emission spectrum of the isotope, the image I k represents the spatial distribution of the activity of the isotope. Usually, I k corresponds to the estimation of the radiation distribution after considering at least one gamma image acquired by the gamma camera. Each item of the image I k (x,y) corresponds to the estimated emissivity or activity of gamma (or X-ray) photons at the point (x,y) in the object plane. The size of the image I k is Nx, Ny, where Nx and Ny represent the number of pixels of the gamma camera along the horizontal axis and the vertical axis, respectively. The size of the matrix I k is the same as the size of the gamma image acquired by the gamma camera.

[0094] F — is the inverse projection operator that allows going from the detection plane P to the object plane P o .

[0095] Indicating the position of the radiation source in the two-dimensional spatial distribution (or mapping) allows simply superimposing the reconstructed image on the visible light image acquired by the visible light camera.

[0096] This method allows locating the gamma source in the environment while allowing the measuring device 1 to move relative to the radiation source. The main steps of this method are in Figure 4is schematically shown in. Generally, the device 1 moves relative to the environment, and during the movement, multiple gamma images are acquired. Between two consecutive acquisitions, the relative movement of the object reference frame relative to the device 1 is recorded. The aim is to allow tracking the position of the radiation source relative to the measurement device 1 in the object plane. The relative movement can be the movement of the measurement device relative to the radiation source, with the radiation source remaining stationary in the environment. It can also be a matter of the movement of the radiation source relative to the measurement device, with the measurement device remaining stationary in the environment.

[0097] Step 100: Initialization

[0098] For the first acquisition, the measurement device 1 is placed at a first position at a first time t1. The gamma image M1 is acquired. Generally, the acquisition time for each gamma image is from 1 ms to 5 s, and preferably from 50 ms to 500 ms, and for example, 100 ms. In the gamma image, each radiation source present in the field takes the form of a trajectory, the brightness of which depends on the radiation generated by the radiation source and detected by the gamma camera.

[0099] When the gamma camera is capable of performing a spectrometry function, as described above, the gamma image is acquired in a given energy spectral band ΔE or a combination of energy bands.

[0100] A visible light image V1 is acquired at the first time t1 or at a time considered consecutive to the first time t1. The main purpose of the initial visible light image V1 is to obtain notable points in the observed visible scene. The notable points and their use are described in more detail in step 130.

[0101] The gamma image M1 is divided by the estimated image The estimated image corresponds to the estimate of the gamma image M1. For the first acquisition, the estimated image is a predetermined image. For example, it is uniform and consists of 1s.

[0102] The ratio corresponds to the error term in the measurement relative to the estimate. For each pixel (u, v) of the gamma image M1 and its estimate the ratio U1 is calculated term by term.

[0103] Step 110: Backpropagation of the error and update of the reconstruction

[0104] The error term U1 is propagated to the object plane in order to update the reconstructed image I1 such that:

[0105] W1(x, y) = F - [U1(u, v)] (2)

[0106] And

[0107] I1(x, y) = I0(x, y) × W1(x, y) (3)

[0108] The image I0(x, y) is a predetermined initialization image, which, for example, only contains positive real numbers, such as just 1.

[0109] After steps 100 to 110, steps 120 to 190 are iteratively executed. Each iteration corresponds to an iteration level k. k is an integer from 2 to K. K corresponds to the iteration level when the iteration stops.

[0110] Step 120: At time t k Acquire the gamma image M k and the visible light image V k .

[0111] At time t k Acquire the visible light image V k . Also at time t k or at an acquisition time considered to be simultaneous with time t k acquire the gamma image M k . Preferably, the acquisition time of each gamma image M k is the same in each iteration k.

[0112] Step 130: Estimation of the motion field D k of

[0113] The purpose of this step is to form a motion field D representing the two-dimensional motion of the visible light image V k-1 relative to the visible light image V k . The image V k is the visible light image acquired in the previous iteration or in step 100 when k = 2. The motion field D k-1 includes motion vectors d corresponding to the motion along the X-axis and the motion along the Y-axis at each coordinate (x, y) in the object plane k . Each motion vector is a vector with a magnitude equal to 2. Therefore, only two-dimensional motion in a plane parallel to or coinciding with the object plane P k is considered. o or coinciding with the object plane is considered.

[0114] The device 1 may have moved or been reoriented between the acquisitions of the images V k-1 and V k . Alternatively, some elements of the object space may have moved between the two acquisitions. In particular, this is the case when the radiation source moves in the object reference system.

[0115] In each image V k-1 and V kAmong them, notable points are identified. The notable points are those that can be easily identified through conventional image processing. For example, this is a matter of points that form the contour or edge of an object or points that have a particularly high contrast from the perspective of brightness or color. Therefore, the notable points can correspond to points of a high Laplacian operator or brightness gradient.

[0116] In each image V k-1 and V k The number of notable points detected is preferably in the dozens to hundreds, or even greater than 1000. In the two images V k-1 and V k The notable points detected respectively form a set E k-1 and a set E k .

[0117] The notable points in the image can be detected by implementing an algorithm (such as the Harris corner detector). After detecting them, the notable points are characterized to allow possible identification in the two images V k-1 and V k . This characterization aims to characterize each notable point and its environment. This can be achieved using feature description algorithms known to those skilled in the art, such as DAISY or LUCID or FREAK. Using the descriptor vector associated with each notable point allows it to be identified in the two images V k-1 and V k .

[0118] Based on their descriptions and their characterizations, the notable points identified in the images V k-1 and V k are matched to establish paired notable points, each pair associating a notable point in the image V k-1 with a notable point in the image V k , and the matched notable points have descriptor vectors that are considered to be the same. The matched notable points correspond to the same point in the observed scene, which appears in both the visible light images V k-1 and V k .

[0119] Preferably, the notable points of each image are considered to be coplanar: they lie in a plane parallel to the plane of the object.

[0120] Based on the matched notable points, the images V k-1 and V k are respectively based on the sets E k-1 and E kMeshing. Delaunay meshing can be used, and this type of meshing generates a triangular mesh cell where each mesh cell lies between three notable points that are different from each other and match in two images V k-1 and V k For each vertex of the mesh existing in images V k-1 and V k the two-dimensional motion d k is estimated. For the preferred embodiment of the present invention, preferably, the motion of the device or the element forming the object reference system is relatively slow so as to maximize the number of vertices of the mesh existing in images V k-1 and V k Sufficiently slow and smooth motion and sufficient light levels also prevent blurring of the visible light image. Preferably, the field of view before and after the motion overlaps by at least 50%, or even at least 80% or 90%.

[0121] Therefore, for each vertex of the mesh existing in images V k-1 and V k an estimate of the motion vector d k is obtained. Since the motion vector is defined only at the vertices existing in images V k-1 and V k interpolation, such as linear interpolation, is performed to obtain an estimate of the motion d k at each point (x, y) of the field of view Ω k of the device, at least within the mesh. More precisely, at least the visible light images V k-1 and V k acquired at measurement times k - 1 and k respectively, the motion is determined in the mesh established in the intersection of the fields of view Ω k-1 and Ω k corresponding thereto. Then, a motion field D k is formed, each item of which corresponds to the motion vector d k (x, y) determined at least at each point (x, y) included in the mesh.

[0122] Based on the motion vector d k (x, y) established within the mesh, the motion field D k outside the mesh can be estimated by extrapolation. This allows obtaining the motion field vector D k for the entire field of view of the image. In one embodiment, the device includes a motion sensor 5 that allows the motion of the device in the visible light images V k-1 and V kEstimate the angular motion between the acquisitions. The motion sensor can be a magneto-inertial measurement unit or include at least one gyroscope. In this case, the motion field outside the grid can be estimated by combining the motion field obtained inside the grid with the motion of the device between two images.

[0123] Other methods allow the estimation of the motion field D between two consecutive images k . For example, this can be a matter of an optical flow method that allows the estimation of the small displacements between two consecutive images.

[0124] By observing the motion of the objects present in the visible light images V k-1 and V k to determine the motion field D at least in part k . According to the first possibility, the object remains stationary in the observed scene: the motion field is established by observing the motion of the object or a part of the object present in the images V k-1 and V k . In this case, the motion field corresponds to the motion of the camera relative to the objects present in the observation field of the camera. According to the second possibility, even if the device has not been moved between the acquisitions of the images V k-1 and V k , the object moves. In this case, the motion field D k corresponds to the motion of the object relative to the visible light camera. More generally, the motion field D k corresponds to the relative motion of the objects present in the observation field of the visible light camera relative to the visible light camera.

[0125] When the motion field outside the grid cannot be determined by extrapolation or using motion measurements, the motion field outside the grid is considered to be uniform and have a constant value, for example, a zero vector can be used.

[0126] At the end of this step, a motion field D of size (2, Nx, Ny) is obtained k . At each relevant point (x, y), a field D of size (Nx, Ny) representing the motion along the X-axis is established k (X) and a field D of size (Nx, Ny) representing the motion along the Y-axis k (Y) .

[0127] Step 140: Estimation of the reconstructed image of

[0128] Based on the reconstructed image I k-1 (x, y) obtained from the previous iteration or initialization and knowing the motion field D k , the reconstructed image is estimated taking into account the registration function This step is an important element of the present invention.

[0129] This step is based on the assumption that the position of the radiation source between times k-1 and k can be reconstructed by considering the motion field D measured between two consecutive visible light images associated with times k-1 and k respectively k to obtain a reconstructed image of the position of the radiation source between times k-1 and k in the motion, visible light images V k-1 and V k are respectively associated with gamma images M k-1 and M k respectively.

[0130] Therefore:

[0131] where:

[0132] × is the element-by-element multiplication operator (Hadamard product);

[0133] I k-1 corresponds to the reconstructed image obtained from the previous iteration. This is a problem of a matrix of the same size as the gamma image M k and each point of this matrix corresponds to the emissivity in a given spectral band or combination of spectral bands, or to the activity of an isotope.

[0134] B is a filter that allows smoothing of the image I k-1 . It can be a low-pass filter, for example, a Gaussian filter. The convolution of the image I k-1 and the smoothing filter B allows the reconstructed image to become blurred. With the help of this filter, it is possible to reconstruct a prior of the position of the radiation source such as described by the reconstructed image I k-1 . According to an alternative, the filtering can be performed according to the following expression:

[0135]

[0136] corresponds to the registration function; is a matrix function of the registration function that takes into account the motion in the object plane while considering the registration of the visible light camera and the gamma camera .

[0137] is the reconstructed image I k-1 after applying the motion vector of the registration.

[0138] At each point (x,y) in the field of view, D k allows defining the motion D along the X axis k (X) (x,y) and the motion D along the Y axisk (Y) (x, y). Matrix function Allows for the motion D k to establish the motion in the object plane. It allows for the estimation of the coordinates (x k-1 , y k-1 ) at time k based on the coordinates (x k , y k ) at time k - 1, such that:

[0139]

[0140] where g X and g Y correspond to the components of the registration functions established along the X-axis and along the Y-axis, respectively.

[0141] Therefore, is the matrix function that allows for the passage between (x k-1 , y k-1 ) and (x k , y k ), as described, for example, by expressions (5.1) and (5.2). This is a problem of variable transformation.

[0142] The exponent α is strictly positive and strictly less than 1: 0 < α ≤ 1. The exponent α allows for the implementation of a memory effect: the closer α is to 1, the more the memory of the previous gamma image acquisition is retained.

[0143] By observing the motion of the object in the observation field of the visible light camera between the visible light images V k-1 and V k , the motion field D k is established. Therefore, the motion of the radiation source in the two reconstructed images I k-1 and is based on the detection of notable points formed and matched in the visible light images V k-1 and V k , and the corresponding motion between them is calculated in the two visible light images. The motion therein allows for the establishment of the motion field D k in the visible light image after applying the registration function, and this motion field is assigned to the gamma image.

[0144] Therefore, within the registration function, the motion of the radiation source is considered to be determinable from the motion field calculated based on the visible light image. The important point in the present invention is that the motion of the object observed in the visible light image is considered to represent the motion of the gamma source, which is shown in the reconstructed image. This greatly limits the computational complexity and the resources required to perform the calculations. The method does not require the implementation of complex techniques such as triangulation or the estimation of the three-dimensional position of the radiation source in the environment.

[0145] Step 150: Gamma image Estimation at time k

[0146] The estimated reconstructed image obtained in step 140 is projected into the detection plane to estimate the gamma image corresponding to iteration k The gamma image is estimated using the following expression

[0147] where F corresponds to the projection operator describing the projection between the object plane P o and the detection plane P

[0148] Step 160: Determination of the error term

[0149] In this step, based on the image M acquired at time k and obtained from step 120 k , and the estimate obtained from step 140 the measurement error is calculated as

[0150] Step 170: Backpropagation of the error

[0151] The error term U k is backpropagated to the object plane using the following expression

[0152] W k (x, y) = F - [U k (u, v)] (8)

[0153] Step 180: Update of the reconstructed image

[0154] In this step, the reconstructed image corresponding to time k is updated according to the expression

[0155] I k = I k-1 × W k (9)

[0156] where × denotes element-by-element multiplication

[0157] According to one embodiment, a non-linear function h is considered, thus

[0158] I k = I k-1 × h(W k ) (10)

[0159] Thus, at each measurement time k, the reconstructed image I is updated by considering the gamma image M acquired at that measurement time k k ​。The error W by backpropagation k to consider the gamma image M k 。

[0160] In the object plane P o and the observation field Ω at the measurement time k k Limit the reconstructed image I in the intersection between them k 。Preferably, the reconstructed image I is not limited outside the observation field Ω at the measurement time k k 。By not considering the radiation sources potentially existing outside the observation field Ω k This allows simplifying the reconstruction process. Then, the reconstructed image only contains the positions of the radiation sources existing in the observation field at the measurement time k, which are obtained based on the gamma image M acquired at the measurement time k or the gamma image acquired at a previous time. In each movement of the measuring device 1, the spatial extent of the reconstructed image is modified so as to limit it to the intersection between the object plane P k and the observation field Ω at the measurement time k O 。 k 。

[0161] The size of the reconstructed image can be less than or equal to the size of each gamma image

[0162] Step 180: Superposition of images

[0163] The reconstructed image I k is given a certain degree of transparency and then superimposed on the visible light image V k without considering the registration function. This allows achieving the correspondence between the objects observed in the visible light image and the radiation sources shown in the reconstructed image

[0164] Step 190:

[0165] Repeat steps 130 to 190 while increasing the iteration index k

[0166] According to one embodiment, the device includes a motion sensor 5, such as an inertial measurement unit. The motion sensor includes a gyroscope, optionally supplemented by an accelerometer and / or a magnetometer. Generally, the motion sensor allows obtaining the angular motion between two consecutive acquisition times. The information obtained from the inertial measurement unit can be considered in step 130 in order to limit the risk of error in the matching of the notable points. It can also be used to complete the motion field D outside the grid formed between the notable points of the matching k 。The translational motion of the camera and the positioning of the object can also be estimated using other devices, such as a LIDAR, a GPS positioning system, a fixed radio frequency beacon, or an inertial navigation system

[0167] According to one embodiment, a gamma camera is capable of performing a spectrometric function. Steps 100 to 180 are performed simultaneously in each energy band ΔE. Then, as many reconstructed images as the energy bands considered can be obtained.

[0168] Steps 100 to 180 can also be performed based on gamma images acquired separately in the respective energy bands and then combined as described above in order to take into account the emission spectrum of a preselected radioisotope. In this case, the reconstructed images can be likened to the spatial distribution of the activity of the radioisotope in the field of view.

[0169] The present invention applies to various nuclear installations or, more generally, to operations for searching for and characterizing radiation sources.

Claims

1. A method for forming a reconstructed image I k wherein the reconstructed image shows the positions (x, y) of radiation sources (10 a , 10 b ) in the observation field Ω, and the method uses a device (1) comprising a gamma camera (2) connected to a visible light camera (3), the device being such that: - The gamma camera and the visible light camera define the observation field Ω; - The visible light camera (3) is configured to form a visible light image V of the observation field Ω k ; - The gamma camera includes pixels configured to detect ionizing electromagnetic radiation generated by a radiation source potentially present in the observation field, and the pixels are located in a detection plane P; - The gamma camera is configured to form a gamma image M k , allowing the estimation of the position of the radiation source in the observation field, the radiation of which is detected by the pixels; The method includes the following iterative steps: a) At an initial measurement time k = 1, acquire an initial gamma image M1 with the gamma camera and acquire an initial visible light image V1 with the visible light camera; b) Initialize a reconstructed image I1 based on the initial gamma image and the initial visible light image; c) Acquire a gamma image M with the gamma camera at measurement time k k and acquire a visible light image V with the visible light camera k ; d) the visible light image V at the measurement time k is compared with the visible light image V at the previous time k - 1, where the previous time is the initial measurement time or the previous measurement time, and then, based on the comparison, the motion field D between the visible light images respectively acquired at the previous time and the measurement time is estimated k-1 ; k ; e) Update the reconstructed image I according to the following k : - The reconstructed image I established at a previous time k-1 ; -The sports ground D obtained in step d) k ; - and the gamma image M acquired at the measurement time k ; f) Optionally, superimpose the reconstructed image or a part of the reconstructed image on the visible light image; g) Repeat steps c) to f) while increasing the measurement time until the iteration is stopped; Among them, The method comprises: moving the device between at least two measurement times such that an observation field Ω k corresponding to each measurement time k - Wherein, at each measurement time, each pixel that has detected radiation generated by a radiation source is associated with a position (x, y) on an object surface arranged facing the detection plane; - Reconstructed image I k corresponds to the spatial distribution of radiation sources on the surface of an object within the observation field Ω k at the measurement time k.

2. The method according to claim 1, wherein The object surface is an object plane arranged parallel to the detection plane P, and each position (x, y) is coplanar.

3. The method according to claim 1, wherein, Observation field Ω only at measurement time k k with the object surface P o the reconstructed image I at the measurement time is defined in the intersection k .

4. The method according to claim 1, wherein At each measurement time, step e) includes: -i) Based on the reconstructed image I established at a previous time k-1 and the motion field D obtained in step d) k , estimate the reconstructed image at the measurement time -ii) Based on the estimated reconstructed image Estimate the gamma image at the measurement time -iii) The gamma image estimated at the measurement time is compared with the gamma image M acquired at the measurement time k ; -iv) Update the reconstructed image I at the measurement time based on the comparison k .

5. The method according to claim 1, comprising: Determine a registration function representing the spatial offset between the gamma camera (2) and the visible light camera (3) The registration function is used in step e).

6. The method according to claim 1, wherein Step d) includes - Select notable points in the visible light images acquired at the previous time and the measurement time respectively; - such that the sports ground D k respectively includes two-dimensional motions of the selected notable points.

7. The method according to claim 6, wherein, The notable points are considered to belong to the same surface.

8. The method according to claim 1, wherein Step d) includes: - Extract a first notable point from the visible light image V acquired at the measurement time k ; - Extract second notable points from the visible light image V acquired at a previous time k-1 ; - Match the first notable point with the second notable point to form paired matching points, each pair being formed by the first notable point and the second notable point; - For each pair of matching points, determine a motion d k (x,y); - Obtaining the motion field D based on the motion determined for each pair of matching points k .

9. The method according to claim 8, including: - Generate a first grid of the observation field at the measurement time using the first notable points, and the first notable points form the first vertices of the first grid; - Generate a second grid of the observation field at the previous time using the second notable points, and the second notable points form the second vertices of the second grid; - Determine the motion at points belonging to the first grid and the second grid by interpolation according to the motion determined for each pair of notable points; - Determine the motion field D based on the motion of each matched vertex and the inserted motion k 。 10. The method according to claim 1, wherein: - The device is coupled to a motion sensor to estimate the motion of the device between two consecutive times; - Step d) takes into account the estimated motion to estimate or verify the motion field.

11. The method according to claim 1, wherein The gamma camera is configured to simultaneously acquire gamma images in various energy bands in steps a) and c), and steps b) and d) to f) are respectively implemented for each energy band so as to obtain reconstructed images in various energy bands at each measurement time.

12. The method according to claim 1, wherein The gamma camera is configured to simultaneously acquire gamma images in various energy bands in steps a) and c), and the method includes acquiring a linear combination of various gamma images in various energy bands at the same measurement time so as to obtain a combined gamma image, and performing steps b) and d) to f) based on the combined gamma image formed at each measurement time.

13. The method according to claim 12, wherein, The combined image is formed from a weighted sum of gamma images acquired at various energies, the weighted sum using weighting factors determined based on the emission spectrum of the radioisotope such that each reconstructed image represents the spatial distribution of the activity of the isotope in the observation field corresponding to the measurement time.

14. A measuring device (1), comprising: - a gamma camera (2) including pixels defining a gamma observation field Ω2, the gamma camera being configured to determine the position of a source of X-ray or gamma radiation in the gamma observation field Ω2; - a visible light camera (3) allowing the formation of a visible light image of a visible light observation field Ω3; - the intersection of the gamma observation field and the visible light observation field is non-zero and defines the observation field Ω of the device; - a processing unit (4) configured to receive at respective measurement times k: · a gamma image formed by the gamma camera; · a visible light image formed by the visible light camera; the processing unit being configured to perform steps b) and d) to g) of the method according to claim 1.

Images