Search

CN-122016902-A - Material identification method based on photon counting X-ray spectrum

CN122016902ACN 122016902 ACN122016902 ACN 122016902ACN-122016902-A

Abstract

The embodiment of the invention provides a material identification method based on photon counting X-ray spectrum, and relates to the technical field of material identification technology. The method comprises the steps of obtaining an original energy spectrum vector of a measured object at a current pixel point, wherein the original energy spectrum vector comprises photon count values of a plurality of energy channels, carrying out logarithmic attenuation correction and smoothing filtering processing on the original energy spectrum vector to generate a physical absorption spectrum vector, generating a K-edge coding feature word based on the physical absorption spectrum vector, and determining a material label of the pixel point based on the K-edge coding feature word. The invention solves the problem of realization, and further improves the real-time property of material identification on a high-speed production line.

Inventors

  • JIN KANGHU

Assignees

  • 芝研智能科技(嘉兴)有限公司

Dates

Publication Date
20260512
Application Date
20260206

Claims (10)

  1. 1. A material identification method based on photon counting X-ray spectrum is characterized by comprising the following steps: Acquiring an original energy spectrum vector of a measured object at a current pixel point, wherein the original energy spectrum vector comprises photon count values of a plurality of energy channels; performing logarithmic attenuation correction and smoothing filtering processing on the original energy spectrum vector to generate a physical absorption spectrum vector; generating a K-edge coding feature word based on the physical absorption spectrum vector, wherein the generation process comprises the steps of calculating a first derivative spectrum of the physical absorption spectrum vector, searching a local peak value meeting a preset derivative threshold value in the first derivative spectrum, and mapping energy position information and amplitude information of the local peak value into an integer bit domain with a preset length; And determining the material label of the pixel point based on the K-edge coding feature word, wherein the determining process comprises the step of matching the K-edge coding feature word with a preset material fingerprint database.
  2. 2. The method of claim 1, wherein said performing log attenuation correction on said raw spectral vector comprises: Acquiring a gain correction matrix and a dark noise matrix based on the empty field scanning data; And calculating the logarithmic attenuation value of the original energy spectrum vector based on the gain correction matrix, the dark noise matrix and a preset minimum protection constant.
  3. 3. The method of claim 1, wherein the generating a K-edge encoded feature word based on the physical absorption spectrum vector comprises: dividing the full spectrum range into a plurality of non-overlapping energy tolerance windows; For each energy tolerance window, searching the maximum derivative value in the window range in the first derivative spectrum of the physical absorption spectrum vector; if the maximum derivative value is smaller than the preset derivative threshold value, the bit position domain corresponding to the window is set to be zero; and if the maximum derivative value is greater than or equal to the preset derivative threshold value, quantizing the maximum derivative value based on a preset amplitude quantization factor, and writing the quantized value into a bit field corresponding to the window.
  4. 4. The method of claim 1, wherein determining the texture label for the pixel based on the K-edge encoded feature word comprises: taking the generated K-edge coding feature words as hash keys, and searching in a hardware hash table deployed in an on-chip memory; If the search hits, directly reading the material identification stored in the hash table as the material label; if the search is not hit, the Hamming distance between the K-edge coding feature word and all the reference feature words in the material fingerprint database is calculated, and the material identifier corresponding to the reference feature word with the minimum Hamming distance and smaller than the preset Hamming threshold value is selected as the material label.
  5. 5. The method according to claim 1, wherein the method further comprises: Calculating a first integral value of the original energy spectrum vector in a high energy section and a second integral value of the original energy spectrum vector in a low energy section; calculating a dual energy ratio assist feature based on the first integral value and the second integral value; Wherein, the determining the texture label of the pixel point based on the K-edge encoding feature word further comprises: and when the K-edge coding feature words are successfully matched, verifying whether the dual-energy-ratio auxiliary features are located in a preset tolerance range or not.
  6. 6. A material identification system based on photon counting X-ray spectroscopy, comprising: The spectrum acquisition module is configured to control the photon counting X-ray detector to acquire an original energy spectrum vector of the measured object at the pixel point; The feature generation module is configured to receive the original energy spectrum vector, execute logarithmic attenuation correction and smooth differential operation on the original energy spectrum vector, and generate a K-edge coding feature word according to derivative peak value features; And the material matching module is configured to store a material fingerprint database, match the K-edge coding feature words with the material fingerprint database and output a material label, wherein the feature generation module and the material matching module are arranged in a hardware acceleration circuit.
  7. 7. The system of claim 6, wherein the texture matching module comprises a hash table stored in on-chip random access memory.
  8. 8. The system of claim 6, wherein the generating a K-edge encoded feature word based on the physical absorption spectrum vector comprises: dividing the full spectrum range into a plurality of non-overlapping energy tolerance windows; For each energy tolerance window, searching the maximum derivative value in the window range in the first derivative spectrum of the physical absorption spectrum vector; if the maximum derivative value is smaller than the preset derivative threshold value, the bit position domain corresponding to the window is set to be zero; And if the maximum derivative value is greater than or equal to the preset derivative threshold value, quantizing the maximum derivative value based on a preset amplitude quantization factor, and writing the quantized value into a bit field corresponding to the window.
  9. 9. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program, wherein the computer program is arranged to execute the method of any of the claims 1 to 5 when run.
  10. 10. An electronic device comprising a memory and a processor, characterized in that the memory has stored therein a computer program, the processor being arranged to run the computer program to perform the method of any of the claims 1 to 5.

Description

Material identification method based on photon counting X-ray spectrum Technical Field The embodiment of the invention relates to the field of material detection, in particular to a material identification method based on photon counting X-ray spectrum. Background The traditional X-ray imaging technology mainly relies on the substance to image the total attenuation (i.e. integral intensity) of X-rays, and the method has good effect in distinguishing substances with large density difference (such as metal and plastic), but has no ability when facing substances with similar density but distinct chemical compositions. For example, in pharmaceutical manufacturing, it is desirable to detect trace amounts of high atomic number contaminants (e.g., lead, mercury particles) incorporated into pharmaceutical powders, or to detect the presence of mixed iodine-or barium-containing specialty chemicals in food packaging. The density of these materials may be similar to the high density innocuous components (such as compacted bones or heavy metal packaging), resulting in failure of conventional gray scale-based detection means. To address this problem, photon counting detectors and multi-spectral (or hyperspectral) imaging techniques have evolved. The technology can acquire the energy spectrum vector of each pixel point, and can accurately identify the material theoretically by analyzing the element characteristics (such as K absorption edge of heavy element) of the material. However, in the application scenario of high-speed production lines (for example, the conveyor belt speed is greater than 0.5 m/s), the conventional pixel-by-pixel full spectrum matching algorithm based on Principal Component Analysis (PCA) or Support Vector Machine (SVM) has a computation delay generally in the millisecond level, and cannot meet the severe requirement of the high-speed production line on microsecond level real-time processing. Disclosure of Invention The embodiment of the invention provides a material identification method and a system based on photon counting X-ray spectrum, which at least solve the problem of poor real-time performance in the related technology. According to an embodiment of the present invention, there is provided a material identification method based on photon counting X-ray spectroscopy, including: Acquiring an original energy spectrum vector of a measured object at a current pixel point, wherein the original energy spectrum vector comprises photon count values of a plurality of energy channels; performing logarithmic attenuation correction and smoothing filtering processing on the original energy spectrum vector to generate a physical absorption spectrum vector; generating a K-edge coding feature word based on the physical absorption spectrum vector, wherein the generation process comprises the steps of calculating a first derivative spectrum of the physical absorption spectrum vector, searching a local peak value meeting a preset derivative threshold value in the first derivative spectrum, and mapping energy position information and amplitude information of the local peak value into an integer bit domain with a preset length; And determining the material label of the pixel point based on the K-edge coding feature word, wherein the determining process comprises the step of matching the K-edge coding feature word with a preset material fingerprint database. In an exemplary embodiment, said performing a logarithmic decay correction on said raw spectral vector comprises: Acquiring a gain correction matrix and a dark noise matrix based on the empty field scanning data; And calculating the logarithmic attenuation value of the original energy spectrum vector based on the gain correction matrix, the dark noise matrix and a preset minimum protection constant. In an exemplary embodiment, the generating the K-edge encoded feature word based on the physical absorption spectrum vector includes: dividing the full spectrum range into a plurality of non-overlapping energy tolerance windows; For each energy tolerance window, searching the maximum derivative value in the window range in the first derivative spectrum of the physical absorption spectrum vector; if the maximum derivative value is smaller than the preset derivative threshold value, the bit position domain corresponding to the window is set to be zero; and if the maximum derivative value is greater than or equal to the preset derivative threshold value, quantizing the maximum derivative value based on a preset amplitude quantization factor, and writing the quantized value into a bit field corresponding to the window. In an exemplary embodiment, the determining the texture label of the pixel point based on the K-edge encoding feature word includes: taking the generated K-edge coding feature words as hash keys, and searching in a hardware hash table deployed in an on-chip memory; If the search hits, directly reading the material identification stored in the hash tabl