Particle trajectory tracking method based on Retina algorithm and FPGA hardware acceleration

Abstract

The invention relates to a particle trajectory tracking method based on a Retina algorithm and FPGA (Field Programmable Gate Array) hardware acceleration, which comprises the following steps of: detecting by using a detector to obtain sample information, inputting the sample information into an FPGA, dividing a two-dimensional space into space units according to a set granularity, and constructing a retina cell calculation array on the FPGA by taking the space units as cell units in a retina algorithm; calculating the similarity weights of the coordinate position information of the particles and the cell units in sequence, and summing the similarity weights to obtain similarity weight values of the cell units; comparing and finding the cell unit with the maximum similarity weight value and the corresponding reconstruction parameter based on a comparator module; the firmware algorithm result output and transmission module outputs the reconstructed particle trajectory to a PC or other terminals; rapid and real-time track tracing and parameter reconstruction can be carried out on linear particle track coordinate position information data.

Description

Technical Field

[0001] The present invention relates to the field of particle trajectory tracking, and in particular to a particle trajectory tracking method based on Retina algorithm and FPGA hardware acceleration. Background Art

[0002] With the rapid development of modern physics, in various high-energy physics experiments related to particle detection, in order to find target events and particles with smaller structures, the energy of the source particles used in the collision or targeting experiments is getting higher and higher. This requires the matching detectors and readout electronics systems to be able to find the target events and particles of interest in a shorter time (tens of μs to several ns). In order to meet such demanding experimental requirements, online real-time particle tracking algorithms are introduced into the readout electronics system and widely used in the field of particle detection technology.

[0003] In recent years, there has been an increasing number of studies on online particle tracking algorithms. Online algorithms are generally placed in the middle and front ends of detector DAQ systems. The research involves the deployment of tracking algorithms on a variety of platforms such as commercial CPUs, GPUs, and FPGAs. Placido Fernandez Declara proposed Compass, a configurable particle tracking algorithm designed for highly parallel architectures and optimized for GPU deployment. Compared with the LHCb baseline, the algorithm's data throughput has increased by 7.4 times; S. Citraro proposed a high-performance "pattern matching" implementation method based on an associative memory system.

[0004] How to quickly identify particle events of interest while reducing the amount of read data, thereby reducing the data transmission bandwidth of the DAQ system and reducing the implementation cost of the readout electronics system is the current research direction. Summary of the invention

[0005] The present invention aims at the technical problems existing in the prior art, provides a particle trajectory tracking method based on Retina algorithm and FPGA hardware acceleration, designs an online trajectory recognition algorithm for SEE (single event effects) positioning, uses the Retina algorithm as a trajectory tracing algorithm, improves the algorithm and combines it with a front-end particle detector system based on a pixel chip to design a dedicated FPGA algorithm firmware to complete the rapid tracing of heavy ion trajectories, realize the physical goal of real-time tracing of heavy ion trajectories in SEE positioning, and can realize rapid and real-time trajectory tracing and parameter reconstruction of the coordinate position information data of the straight line particle trajectory detected by the pixel detector in the two-dimensional plane space, thereby completing the positioning task of the single particle effect in the two-dimensional space.

[0006] According to a first aspect of the present invention, a particle trajectory tracking method based on Retina algorithm and FPGA hardware acceleration is provided, comprising: Step 1, using a detector to detect and obtain sample information and input the sample information into an FPGA; the sample information includes: coordinate position information of pixel points when particles pass through each layer of the detector plane; Step 2, dividing the two-dimensional space into various spatial units according to the set granularity, and using the spatial units as the cell units in the retinal algorithm to construct a retinal cell computing array on the FPGA; calculating the similarity weights of various coordinate position information of the particles and various cell units in sequence based on the longitudinal distance, and summing the similarity weights of each cell unit with each coordinate position information of the particle to obtain the similarity weight value of the cell unit; Step 3, constructing a comparator module on the FPGA, and finding the cell unit and its corresponding reconstruction parameters when the similarity weight value is the largest based on the comparison of the comparator module; Step 4: construct a firmware algorithm result output transmission module on the FPGA, and the firmware algorithm result output transmission module outputs the reconstructed particle trajectory to a PC or other terminals.

[0007] Based on the above technical solution, the present invention can also make the following improvements.

[0008] According to a second aspect of the present invention, there is provided a particle trajectory tracking device based on the Retina algorithm and FPGA hardware acceleration, comprising: a sample event storage and input module and an FPGA; The sample event storage and input module is used to obtain sample information detected by the detector and input the sample information into the FPGA; the sample information is the coordinate position information of the pixel points when one or more particles pass through each layer of the detector plane; The FPGA includes: a Retina retinal cell calculation array, a comparator module and a firmware algorithm result output transmission module; Divide the two-dimensional space into various spatial units according to the set granularity, and use the spatial units as cell units in the retina algorithm to establish the Retina retina cell calculation array, wherein the retina cell calculation array is used to sequentially calculate the similarity weights of various coordinate position information of the particle and various cell units based on the longitudinal distance, and each of the cell units is summed up with the similarity weights of various coordinate position information of the particle to obtain the similarity weight value of the cell unit; The comparator module is used to find the cell unit and its corresponding reconstruction parameter when the similarity weight value is the largest based on the comparison of the comparator module; The firmware algorithm result output transmission module is used to output the reconstructed particle trajectory to a PC or other terminals.

[0009] According to a third aspect of the present invention, an electronic device is provided, comprising a memory and a processor, wherein the processor is used to implement the steps of a particle trajectory tracking method based on the Retina algorithm and FPGA hardware acceleration when executing a computer management program stored in the memory.

[0010] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which a computer management program is stored. When the computer management program is executed by a processor, the steps of a particle trajectory tracking method based on the Retina algorithm and FPGA hardware acceleration are implemented.

[0011] The present invention provides a particle trajectory tracking method, device, electronic device and storage medium based on Retina algorithm and FPGA hardware acceleration, and the position resolution is When the algorithm granularity is When the firmware global clock frequency is 50MHz, the algorithm takes 49 system clock cycles or 0.98μs to process a single position information in a single heavy ion trajectory sample event. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flow chart of a particle trajectory tracking method based on Retina algorithm and FPGA hardware acceleration provided by the present invention; Figure 2 A structural block diagram of a particle trajectory tracking device based on Retina algorithm and FPGA hardware acceleration provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of an embodiment of the Retina retinal cell computing array provided by the present invention; Figure 4 A schematic structural diagram of an embodiment of a comparator module provided by the present invention; Figure 5 A schematic diagram of the hardware structure of a possible electronic device provided by the present invention; Figure 6 A schematic diagram of the hardware structure of a possible computer-readable storage medium provided by the present invention. DETAILED DESCRIPTION

[0013] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0014] Example 1 Embodiment 1 provided by the present invention is an embodiment of a particle trajectory tracking method based on Retina algorithm and FPGA hardware acceleration provided by the present invention. Figure 1 A flow chart of particle trajectory tracking based on Retina algorithm and FPGA hardware acceleration is provided in an embodiment of the present invention, such as Figure 1 As shown, an embodiment of the method includes: Step 1: Use a detector to detect and obtain sample information and input the sample information into FPGA; the sample information includes: coordinate position information of pixel points when particles pass through each layer of detector plane.

[0015] In the specific implementation process, the single track sample information used for simulation sample input comes from the Topmetal pixel chip, which is a semiconductor detector installed in the heavy ion detection platform. The closed gas chamber inside the platform is filled with inert gas. When the charged particle passes through the gas detector, the ions and photons along the particle flight track will react with the inert gas in the device to ionize electron-hole pairs. Under the action of the external electric field, the electrons drift toward the Topmetal pixel array as the anode, and finally form a projected trajectory image of the charged particle. The principle of using the Topmetal sensor to track a single charged particle is shown in the figure. The positioning module consists of two positioning units. In each positioning unit, the silicon pixel sensor acts as an anode to collect drifting electrons for tracking imaging. The two silicon pixel sensors are placed at appropriate positions on two orthogonal planes. Through position correction, the two-dimensional position information of the incident charged particle can be obtained. When the sample is generated, 25MeV / u Kr ions are used as incident charged particles, passing through a gas chamber filled with 70:30 argon and carbon dioxide at a pressure of 300K, with an electric field strength of 300V / cm. The pixel array of the Topmetal sensor is , a single pixel size of 83μm, each sensor provides The effective scanning area.

[0016] In a possible embodiment, in order to realize the real-time tracking of a single heavy ion event, the sample constructed by the embodiment of the present invention is 5000 frames of single-track particle trajectory data generated by the Topmetal sensor, each frame contains a resolution of The two-dimensional position distribution of the particle flight trajectory, that is, 5184 coordinate information, is output in the form of coordinate points. Only the integer part of the coordinate point value is retained, and the value range is 1 to 72.

[0017] In order to better evaluate the tracking accuracy of the algorithm, the embodiment of the present invention also introduces two new parameters: distance threshold and pixel weight value. Pixel value weight refers to the energy value of each pixel on the pixel sensor, which is determined by the pixel being hit by electrons, and the pixel value weight value retains 4 decimal places; distance threshold refers to setting a pixel value weight to determine whether to use the pixel's position information as an input sample. Only the position information of pixels whose pixel value weight is greater than the distance threshold will be used as sample coordinate data input to the retinal array. At the same time, the pixel value weight will participate in the calculation of the retinal unit, and provide appropriate weights to the contribution value provided by each input element to help screen those more valuable pixel position information. Each frame of samples processed by threshold filtering can generate 200-300 coordinate point data and pixel value weights. These coordinate and pixel value information are called a set of single track sample events. According to statistics, the number of coordinate points of 5000 frame sample events is between 160 and 240, and the average value is about 210.

[0018] Step 2: Divide the two-dimensional space into various spatial units according to the set granularity, and use the spatial units as the cell units in the retinal algorithm to build a retinal cell computing array on the FPGA; calculate the similarity weights of each coordinate position information of the particle and each cell unit in turn based on the longitudinal distance, and sum up the similarity weights of each cell unit with the various coordinate position information of the particle to obtain the similarity weight value of the cell unit.

[0019] The retinal algorithm is inspired by the recognition characteristics of biological retina and is a pattern recognition algorithm invented by imitating the biological visual system.

[0020] The first step is to construct the parameter space, that is, the pattern set. The two-dimensional Retina parameter space is established with two parameters, the vertical intercept b and the slope k of the straight line trajectory in the two-dimensional space. After limiting the value range of k and b, a straight line y=kx+b in the physical space corresponds to a coordinate point (k, b) in the parameter space.

[0021] The second step is to divide the established Retina parameter space into equal parts according to a certain granularity to form sub-parameter spaces. A single sub-parameter space is called a retinal computing unit, also called a retinal cell unit (Retinacell). In this way, the entire parameter space is The retinal cell array consists of retinal cell units. The center coordinate (i, j) of the parameter space where each retinal cell unit is located corresponds to a straight line template y=ix+j in the physical space. In the retinal algorithm, the granularity of the algorithm is It is used to measure the density of pattern units contained in the parameter space and is an important parameter for debugging the algorithm.

[0022] The third step is to input the straight line sample information into a single retinal cell (i, j) to calculate the similarity weight value of the sample information .

[0023] In a possible embodiment, the formula (1) for obtaining the similarity weight value of the cell unit based on the longitudinal distance is:

[0024] Where i and j represent the horizontal and vertical serial numbers of the cell unit / super retinal unit, respectively; m and k represent the total number and serial number of the layers of the parallel plane of the detector, respectively; is the similarity weight value of the coordinate position information of the particle detected by the cell unit / super retinal unit in the i-th row and j-th column and the k-th layer detector; The coordinate position information of the particle detected by the k-th layer detector and the distance between the cell unit / super retinal unit in the i-th row and j-th column in the longitudinal direction; It is the adjustment factor of the operation, and its size is determined by the granularity of the retina.

[0025] Step 4: Calculate the similarity weight value according to step 3 The method scans all retinal cells in the retinal unit array and obtains the input sample in the retinal parameter space. The distribution of values. In general, the algorithm will The parameter space position (i_sum max, j_sum max) corresponding to the cell unit with the largest value is output as the identified straight line trajectory and the reconstructed longitudinal intercept and slope parameters.

[0026] In a possible embodiment, in order to better describe how to use the retina algorithm to track linear particle trajectories in two-dimensional space, the process of the retina algorithm is simulated using Python. A parameter space (k, b) is set in the two-dimensional space, where the value range of k is (-1, 1), representing the slope of the particle trajectory reconstructed by the algorithm; the value range of b is (0, 72), representing the intercept of the reconstructed particle trajectory. The granularity is set to , i.e. 4900 retinal cell units. Then a particle track represented by y=0.025k+34 in two-dimensional space (x, y) was simulated, and 210 coordinate points on the track were taken as sample event elements and input into the array, resulting in Distribution, the parameter space coordinates corresponding to the retinal unit with the largest similarity weight value are (0.014, 34.35), that is, the straight line trajectory parameter recognized by the retinal array is y=0.014x+34.35.

[0027] Step 3: construct a comparator module on the FPGA, and compare and find the cell unit with the largest similarity weight value and its corresponding reconstruction parameters based on the comparator module.

[0028] Step 4: construct a firmware algorithm result output transmission module on the FPGA, and the firmware algorithm result output transmission module outputs the reconstructed particle trajectory to a PC or other terminal.

[0029] In a possible implementation manner, step 4 further includes: A netlist configuration module is set in the FPGA. The netlist configuration module calculates the parameters of any cell unit in the i-th row and j-th column based on the input granularity parameters M and N. .

[0030] in, ; .

[0031] M and N represent the number of horizontal and vertical cell units in the retinal cell calculation array, respectively. and They represent the slope and intercept of the cell unit in the i-th row and j-th column respectively.

[0032] In a possible embodiment, the process of determining the straight trajectory of the particle based on the position and outputting it includes: Determine the horizontal and vertical serial numbers i and j of the cell unit / super retinal unit, and use the slope and intercept of the cell unit / super retinal unit in the i-th row and j-th column as the slope and intercept of the particle linear track.

[0033] In SEE positioning, the key issue is to quickly identify the trajectory of heavy ions, the main target particles that produce the SEE effect. The present invention provides a particle trajectory tracking method based on the Retina algorithm and FPGA hardware acceleration, designs an online trajectory recognition algorithm for SEE positioning, uses the Retina algorithm as a trajectory tracking algorithm, improves the algorithm and combines it with a front-end pixel chip-based particle detector system to design a dedicated FPGA algorithm firmware to complete the rapid tracking of heavy ion trajectories, thereby realizing the physical goal of real-time tracking of heavy ion trajectories in SEE positioning.

[0034] Example 2 Embodiment 2 provided by the present invention is an embodiment of the present invention based on the Retina algorithm and FPGA hardware acceleration method, and the two-dimensional parameter space of the retina algorithm is (k, b). Wherein k represents the slope of the straight line, and the value range is set to -1~1. And b represents the vertical intercept of the straight line, and the value range is set to 0~72 pixel size.

[0035] In order to improve the calculation accuracy of the retinal algorithm in the experimental environment, three new influencing factors are added to the link of calculating the weight value W in the retinal algorithm calculation unit: pixel value, pixel threshold and distance threshold. In the specific implementation, three read-only memories can be used to store the sample information, and the three read-only memories are used to store the x-coordinate information, y-coordinate information and pixel brightness value of the valid pixel points in each sample event respectively. The three data are all stored in the ROM with fixed bit width in the form of fixed-point decimals.

[0036] In a possible embodiment, the pixel value An of each input sample coordinate point n is added as an influencing factor to the similarity weight calculation formula (1). The calculation formula becomes as shown in formula (2):

[0037] Using formula (2) can make the coordinate point pair samples with high brightness (pixel value) in the trajectory sample have a higher matching weight in the predicted line L represented by the corresponding visual cell. This can improve the tracking accuracy of the retinal algorithm.

[0038] In a possible embodiment, based on the similarity weight calculation formula (2), a distance threshold d_th and a pixel threshold A_th are further added. The calculation formula becomes as shown in formula (3):

[0039]

[0040] Since a straight line track sample contains a large number of valid coordinate points, formula (3) sets two thresholds d_th and A_th to allow the algorithm to ignore the contribution of coordinate positions that are far away from the predicted straight line L (dij_n>d_th) or have low pixel values ​​An to the weight value W, thereby reducing the amount of calculation of the algorithm while improving the algorithm's calculation efficiency and tracking accuracy.

[0041] According to the generated heavy ion trajectory samples, the pixel threshold value A_th is set to 10 brightness, and the distance threshold value d_th is set to 1.5 times the size of a single retinal computing unit on the vertical intercept parameter axis, that is, , J represents the particle size of the intercept.

[0042] Example 3 Embodiment 3 provided by the present invention is an embodiment of a particle trajectory tracking device based on the Retina algorithm provided by the present invention, such as Figure 2 FIG. 1 is a schematic diagram of a structure of an embodiment of the trajectory tracking device provided by an embodiment of the present invention, Figure 2It can be seen that the embodiment of the trajectory tracking device includes: FPGA. The FPGA includes: a sample event storage and input module, a Retina retinal cell calculation array, a comparator module and a firmware algorithm result output transmission module.

[0043] The sample event storage and input module is used to obtain the sample information detected by the detector and input the sample information into the Retina retinal cell computing array; the sample information is the coordinate position information of the pixel point when one or more particles pass through each layer of the detector plane.

[0044] In a specific implementation, the sample event storage and input module can be composed of three read-only memories, which are used to store the x-coordinate information, y-coordinate information and pixel brightness value of each valid pixel in each sample event. The three data are stored in a fixed-width ROM in the form of fixed-point decimals.

[0045] The two-dimensional space is divided into various spatial units according to the set granularity, and the spatial units are used as the cell units in the retina algorithm to establish a Retina retina cell computing array. The retina cell computing array is used to calculate the similarity weights of each coordinate position information of the particle and each cell unit based on the longitudinal distance in sequence. The similarity weights of each cell unit with the various coordinate position information of the particle are summed to obtain the similarity weight value of the cell unit.

[0046] The Retina retinal cell computing array is responsible for allocating and generating a retinal array of corresponding size according to the input granularity, and then calculating and outputting the similarity weight value of each retinal cell based on the input sample information.

[0047] The comparator module is used to find the cell unit with the largest similarity weight value and its corresponding reconstruction parameter based on the comparison of the comparator module.

[0048] When the sample is a single-particle sample, we use this module to find the maximum similarity weight value and output the corresponding cell unit coordinate information (i, j) as the straight line trajectory found by the algorithm and its corresponding reconstruction parameters.

[0049] The firmware algorithm result output transmission module is used to output the reconstructed particle trajectory to a PC or other terminal. In the embodiment of the present invention, the IPBUS bus is used as the main upload data exchange bus.

[0050] The retina array is a core component of the tracking firmware. In a possible embodiment, it is composed of a netlist and a large number of retina cells. During the firmware design process, a fixed number of retina cell units are placed in the retina array according to the firmware resources of the specific FPGA actually used. When the algorithm is working, after the granularity information is input through the netlist, the number of retina cell units corresponding to the input granularity information is activated to the working state, and the retina cell units corresponding to the excess are suspended and enter the idle state, thereby realizing the online configuration function of the retina array. When the number of retina cell units that need to be activated exceeds the total number of retina cell units pre-solidified into the firmware algorithm, the netlist module uses the repeated configuration function to reuse the retina cell units in the array to complete the scanning of the retina algorithm. In addition, the retina algorithm is integrated into all retina cell units, which can simultaneously perform similarity weighting on the input sample event elements. The calculation ensures the high parallelism of the algorithm, giving the algorithm a speed advantage in real-time particle trajectory tracing applications.

[0051] In the retina array module, the Retina cell unit is the core module of the entire Retina algorithm firmware. The input of the Retina cell unit module is divided into two parts: the first part is the position coordinate data and the corresponding pixel brightness value information from the heavy ion sample; the second part is the parameter pair (ki, bj) assigned to the retina cell unit from the netlist configuration module. The parameter pair assigned from the netlist configuration module is a fixed parameter during the retina cell unit operation, while the heavy particle sample information is input into the retina cell unit module in a serial manner.

[0052] From the overall design framework of the Retina algorithm, we can see that most of the algorithm space is composed of a whole array of retina cell modules. Therefore, the design structure, resource consumption and computational latency of the Retina cell module play a decisive role in the final real-time performance and resource consumption of the Retina firmware algorithm. In order to obtain the optimal algorithm firmware online performance indicators, in the firmware structure design of the Retina cell, it is necessary to select the most optimized structure to balance the two major indicators of computational latency and resource consumption.

[0053] like Figure 3 FIG. 1 is a schematic diagram showing the structure of an embodiment of the Retina retinal cell computing array provided by the present invention, combined with Figure 3 It can be seen that the overall structure of the Retina retinal cell computing array calculates the weight value and The steps are arranged in sequence. In order to optimize the calculation delay and resource consumption, each calculation step is designed with a pipeline structure. Since each calculation step has a different calculation delay, each unit is connected with a fifo to ensure that the calculation result will not be lost. Using the pipeline structure, the retina cell unit can realize the input and calculation of sample point data information without delay. The design goal is to output the value, that is, to realize that the algorithm can complete the input of valid pixel information in a sample and a calculation result every time a system clock cycle passes. Under this condition, in order to reduce the resource consumption of a single retina cell unit module as much as possible, the entire The computing process is divided into multiple minimum delay computing units that cannot be further divided, and the Retina retinal cell computing array includes multiple minimum delay computing units.

[0054] Each minimum delay calculation unit is connected in sequence, and is used to complete the addition and multiplication calculation of the output dij_v, the square calculation of the output dij_v2 value, Multiplication calculation, Lookup table calculation and similarity weight value of cell unit The cumulative calculation of .

[0055] When the pipeline method is used to implement the input and output delay-free computing function, the number of pipelines in the pipeline is equal to the system clock cycle consumed by the corresponding calculation of the module. Combined with the calculation steps of the designed retina cell unit, the calculation steps of the designed single retina cell unit module and its corresponding calculation delay, the function completed by the calculation, and the calculation resource overhead are finally obtained. Among them, the total delay of the retina cell unit calculation is 41 system clock cycles, and the time required to calculate a heavy particle trajectory sample is 41+V system clock cycles, where V is the number of valid pixel information contained in the heavy particle sample, which is between 150-250, with an average of 210. A single Retina cell unit is composed of 2 LUT lookup tables, 5 addition and subtraction calculation units, and 11 multiplication calculation units, consuming a total of 5 DSP48 resources and a small amount of LUT resources.

[0056] like Figure 4 FIG. 1 is a schematic diagram showing a structure of an embodiment of a comparator module provided by the present invention, Figure 4 It can be seen that in a possible embodiment, the comparator module is an 8-level comparator network composed of a plurality of two-input comparators connected in series and parallel in a cascade manner.

[0057] The comparator module is the last step of the Retina algorithm calculation step. Its function is to calculate the value of all retina cells in the retina array. The values ​​are compared and arranged to select the maximum value Summax. When Summax is greater than the given threshold Sum_th, the algorithm successfully finds a heavy particle trajectory, and the parameters (ki, bj) corresponding to the Summax are the reconstructed trajectory parameters. In the embodiment of the present invention, the two-input comparator is connected in series and parallel to form an 8-level comparator network. After 8 clock cycles, the module can complete the tracking result output of a single particle trajectory sample.

[0058] In a possible embodiment, in order to realize the function of online configuration of Retina scan granularity when designing the algorithm firmware, so as to make the algorithm universal, the netlist configuration module is designed and added to the retina algorithm array module. The input parameter of the netlist configuration module is the granularity of the retina algorithm. In the application of the present invention, the parameter space of the algorithm is a two-dimensional parameter space, so the input parameters of the netlist configuration module are M and N. Before the retina array performs calculations, the netlist configuration module receives the granularity parameters M and N. The netlist configuration module is internally designed to consist of a fixed-point DSP48 computing unit and necessary logic functions. After the calculation of the netlist module, the module finally outputs a granularity that matches the input granularity. parameter pairs [ki, bj], which are connected to the corresponding retina cell modules and used as the configuration of the scanning parameters of each retina cell module.

[0059] In the actual design, according to the characteristics of the heavy ion samples generated by the experimental environment mentioned above, the parameter space value range of the particle trajectory slope parameter k is set to -1~1 and the parameter space value range of the vertical intercept b is set to (The size of a single pixel is ). Therefore, the specific calculation formula for the parameter pair ki and bj in the netlist firmware is:

[0060]

[0061] The present invention provides an embodiment of a particle trajectory tracking device based on the Retina algorithm. In the firmware design, the KC705 development board equipped with the Xilinx 7 series FPGA chip is selected as the platform for the firmware design implementation. The firmware algorithm input sample is a single particle sample event of heavy ion Kr read out by the pixel detector array chip Topmetal chip. The pixel array scale is , the size of a single pixel is In the test, retinal algorithms with different granularities were selected and detailed algorithm performance evaluation results were obtained. Under the firmware algorithm conditions, the algorithm's position resolution reached 18.11μm. When the firmware global clock frequency is 50MHz, the time consumed to process a standard single-track heavy ion trajectory sample event with 210 coordinates is 5.18μs. The above test results show that the algorithm can achieve fast and real-time heavy ion trajectory reconstruction with high resolution, thereby achieving real-time SEE positioning.

[0062] It can be understood that the particle trajectory tracking device based on the Retina algorithm provided by the present invention corresponds to the particle trajectory tracking based on the Retina algorithm and FPGA hardware acceleration provided by the aforementioned embodiments. The relevant technical features of the particle trajectory tracking device based on the Retina algorithm can refer to the relevant technical features of the particle trajectory tracking based on the Retina algorithm and FPGA hardware acceleration, which will not be repeated here.

[0063] See also Figure 5 , Figure 5 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 5 As shown, an embodiment of the present invention provides an electronic device, including a memory 1310, a processor 1320, and a computer program 1311 stored in the memory 1310 and executable on the processor 1320. When the processor 1320 executes the computer program 1311, the following steps are implemented: dividing the two-dimensional space into various spatial units according to the set granularity, and establishing a cell array with the spatial units as the cell units in the retinal algorithm; using the detector to detect and obtain the coordinate position information of the pixel point when a particle passes through each layer of the detector plane; calculating the similarity weights of each coordinate position information of the particle and each cell unit in turn based on the longitudinal distance, summing the similarity weights of a cell unit and each coordinate position information of the particle to obtain the similarity weight value of the cell unit, traversing all the cell units in the cell array, determining the position of the cell unit corresponding to the largest similarity weight value, and determining the straight line trajectory of the particle based on the position and outputting it.

[0064] See also Figure 6 , Figure 6 Schematic diagram of an embodiment of a computer-readable storage medium provided by the present invention. Figure 6As shown, this embodiment provides a computer-readable storage medium 1400, on which a computer program 1411 is stored. When the computer program 1411 is executed by a processor, the following steps are implemented: dividing the two-dimensional space into various spatial units according to a set granularity, and establishing a cell array with the spatial units as the cell units in the retinal algorithm; using a detector to detect and obtain the coordinate position information of the pixel point when a particle passes through each layer of the detector plane; calculating the similarity weights of each coordinate position information of the particle and each cell unit in turn based on the longitudinal distance, summing the similarity weights of a cell unit and each coordinate position information of the particle to obtain the similarity weight value of the cell unit, traversing all the cell units in the cell array, determining the position of the cell unit corresponding to the largest similarity weight value, and determining the straight line trajectory of the particle based on the position and outputting it.

[0065] The embodiment of the present invention provides a particle trajectory tracking device, electronic device and storage medium based on Retina algorithm and FPGA hardware acceleration, and the position resolution is within the algorithm granularity of When the algorithm granularity is When the global clock frequency of the firmware is 50MHz, the time consumed by the algorithm to process a single position information in a single heavy ion trajectory sample event is 49 system clock cycles, or 0.98μs. When the calculation part is designed with a pipeline in the retina cell unit, the time consumed by the algorithm firmware to process a single heavy ion sample event is between 3.98μs and 5.98μs, with an average of 5.18μs. A single retina computing unit consumes 5 DSP48 units. When 100 retina computing units are embedded, the entire algorithm firmware consumes 74.63% of the LUT, 41.81% of the flip-flop, 67.41% of the block RAM and 83.32% of the DSP48 resources of the K7-325tFPGA chip.

[0066] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0067] It should be understood by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0068] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0069] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0070] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0071] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0072] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

[0073] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0074] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0075] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0076] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0077] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A particle trajectory tracking method based on Retina algorithm and FPGA hardware acceleration, characterized in that: The particle trajectory tracking method comprises: Step 1, using a detector to detect and obtain sample information and input the sample information into an FPGA; the sample information includes: coordinate position information of pixel points when particles pass through each layer of the detector plane; Step 2, dividing the two-dimensional space into various spatial units according to the set granularity, and using the spatial units as the cell units in the retinal algorithm to construct a retinal cell computing array on the FPGA; calculating the similarity weights of various coordinate position information of the particles and various cell units in sequence based on the longitudinal distance, and summing the similarity weights of each cell unit with each coordinate position information of the particle to obtain the similarity weight value of the cell unit; Step 3, constructing a comparator module on the FPGA, and finding the cell unit and its corresponding reconstruction parameters when the similarity weight value is the largest based on the comparison of the comparator module; Step 4: construct a firmware algorithm result output transmission module on the FPGA, and the firmware algorithm result output transmission module outputs the reconstructed particle trajectory to a PC or other terminals.

2. The trajectory tracking method according to claim 1, characterized in that: The formula for obtaining the similarity weight value of the cell unit based on the longitudinal distance is: Among them, i and j represent the horizontal and vertical serial numbers of the cell unit respectively; m and k represent the total number and serial number of the layers of the parallel plane of the detector respectively; is the similarity weight value of the coordinate position information of the cell unit in the i-th row and the j-th column and the particle detected by the k-th layer detector; The coordinate position information of the particle detected by the k-th layer detector and the distance between the cell unit in the i-th row and j-th column in the vertical direction; is the adjustment factor for the operation.

3. The trajectory tracking method according to claim 1, characterized in that: The sample information also includes: pixel brightness values ​​of the pixels when the particles pass through the detector planes of each layer; The formula for obtaining the similarity weight value of the cell unit / super retinal unit based on the longitudinal distance is: Where i and j represent the horizontal and vertical serial numbers of the cell unit / super retinal unit, respectively; m and n represent the total number and serial number of the layers of the parallel plane of the detector, respectively; is the similarity weight value of the coordinate position information of the cell in the i-th row and j-th column and the particle detected by the detector in the n-th layer; The coordinate position information of the particle detected by the n-th layer detector and the distance between the cell unit / super retinal unit in the i-th row and j-th column in the longitudinal direction; is the adjustment factor of the operation; An represents the pixel brightness value of the pixel point of the particle detected by the n-th layer detector; d_th is the set distance threshold, and A_th is the set pixel threshold.

4. The particle trajectory tracking method according to claim 1, characterized in that: The step 4 also includes: A netlist configuration module is set in the FPGA, and the netlist configuration module calculates the parameters of any cell unit in the i-th row and j-th column based on the input granularity parameters M and N. ; in, ; ； M and N represent the number of horizontal and vertical cell units in the retinal cell calculation array, respectively. and They represent the slope and intercept of the cell unit in the i-th row and j-th column respectively.

5. The trajectory tracking method according to claim 1, characterized in that: The reconstruction parameters corresponding to the cell unit when the similarity weight value is the largest in step 3 include: the horizontal and vertical serial numbers i and j of the cell unit; The process of reconstructing the particle trajectory obtained in step 4 includes: The slope and intercept of the cell unit in the i-th row and j-th column are used as the slope and intercept of the particle reconstructed trajectory.

6. A particle trajectory tracking device based on Retina algorithm and FPGA hardware acceleration, characterized in that: include: FPGA; The FPGA includes: a sample event storage and input module, a Retina retinal cell calculation array, a comparator module and a firmware algorithm result output transmission module; The sample event storage and input module is used to obtain sample information detected by the detector and input the sample information into the Retina retinal cell computing array; the sample information is the coordinate position information of the pixel points when one or more particles pass through each layer of the detector plane; Divide the two-dimensional space into various spatial units according to the set granularity, and use the spatial units as cell units in the retina algorithm to establish the Retina retina cell calculation array, wherein the retina cell calculation array is used to sequentially calculate the similarity weights of various coordinate position information of the particle and various cell units based on the longitudinal distance, and each of the cell units is summed up with the similarity weights of various coordinate position information of the particle to obtain the similarity weight value of the cell unit; The comparator module is used to find the cell unit and its corresponding reconstruction parameter when the similarity weight value is the largest based on the comparison of the comparator module; The firmware algorithm result output transmission module is used to output the reconstructed particle trajectory to a PC or other terminals.

7. The trajectory tracking device according to claim 6, characterized in that: The Retina retinal cell computing array includes a plurality of minimum delay computing units; Each of the minimum delay calculation units is connected in sequence, and is used to complete the addition calculation and multiplication calculation of the output dij_v, the square calculation of the output dij_v2 value, Multiplication calculation, Lookup table calculation and similarity weight value of cell unit The cumulative calculation of .

8. The trajectory tracking device according to claim 6, characterized in that: The comparator module is an 8-level comparator network composed of multiple two-input comparators connected in series and parallel.

9. An electronic device, characterized in that: It comprises a memory and a processor, wherein the processor is used to implement the steps of the particle trajectory tracking method based on the Retina algorithm and FPGA hardware acceleration as described in any one of claims 1 to 6 when executing a computer management program stored in the memory.

10. A computer-readable storage medium, characterized in that: A computer management program is stored thereon, and when the computer management program is executed by the processor, the steps of the particle trajectory tracking method based on the Retina algorithm and FPGA hardware acceleration as described in any one of claims 1 to 6 are implemented.

Images