# US 12,216,228 B1 · Method and device for predicting impact point based on non-linear maneuver identification of flying object > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,216,228 B1 - Document ID: US12216228B1 - Status: Granted - Kind: B1 - Application number: 18/624,997 - Filed: 2024-04-02 - Granted: 2025-02-04 - Earliest claimed: 2024-01-31 - Assignees: Agency for Defense Development - Applicants: Agency for Defense Development - Inventors: Ho Cheol Jeon; Hyoek Jin Choi; Jae Kak Choi; Young Min Kim - Web page: https://patentgrep.com/patent/US12216228B1 ## CPC - G01S 7/415: {Identification of targets based on measurements of movement associated with the target} - G01S 7/417: {involving the use of neural networks} - G01S 13/723: {by using numerical data} - G01S: RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES ## Abstract A method for predicting an impact point is proposed. The method may include acquiring detection information related to a flying object detected by a radar linked to the impact point predicting device through a communication interface, acquiring prior information including information about an enemy unit, a friendly unit, and a friendly facility. The method may include determining whether the detected flying object is a flying object capable of non-linear maneuver based on the detection information and the prior information, using a neural network model pre-trained using training data so that the neural network model determines whether the flying object is the flying object capable of non-linear maneuver based on detection information and prior information and determining an impact point based on the detection information, the prior information, and a determining result of the neural network model on whether the detected flying object is the flying object capable of non-linear maneuver. ## Claims ### Claim 1 A method for predicting an impact point using an impact point predicting device including a processor, the method comprising: acquiring detection information related to a flying object detected by a radar linked to the impact point predicting device through a communication interface; acquiring prior information including information about an enemy unit, a friendly unit, and a friendly facility; determining whether the detected flying object is a flying object capable of non-linear maneuver based on the detection information and the prior information, using a neural network model pre-trained using training data so that the neural network model determines whether the flying object is the flying object capable of non-linear maneuver based on detection information and prior information; and determining an impact point based on the detection information, the prior information, and a determining result of the neural network model on whether the detected flying object is the flying object capable of non-linear maneuver, wherein the neural network model includes an input layer set to input detection information and prior information of a flying object, an output layer outputting the determining result, and a model parameter trained to have a correlation between the detection information, the prior information, and the determining result based on a supervised training by using label information specifying whether the flying object is capable of non-linear maneuver. ### Claim 2 (depends on 1) The method of claim 1, wherein acquiring the detection information includes extracting dynamic characteristic information of the detected flying object by replaying an image of the flying object detected by the radar. ### Claim 3 (depends on 2) The method of claim 2, wherein the dynamic characteristic information contains information about a launch point, moving position, moving trajectory, speed, acceleration, and RCS of the detected flying object. ### Claim 4 (depends on 1) The method of claim 1, wherein acquiring the detection information includes extracting dynamic characteristic information of the detected flying object by replaying an image of the flying object detected by the radar, and wherein the neural network model uses the extracted dynamic characteristic information as training data for the detection information that is to be input into the input layer. ### Claim 5 (depends on 1) The method of claim 1, wherein determining whether the detected flying object is the flying object capable of non-linear maneuver includes assigning some unspecified detection information to a preset value and inputting the unspecified detection information into the neural network model, when some of multiple pieces of detection information to be input into the neural network model are unspecified. ### Claim 6 (depends on 1) The method of claim 1, wherein acquiring the detection information includes updating the detection information at a preset cycle. ### Claim 7 (depends on 6) The method of claim 6, wherein determining whether the flying object is the flying object capable of non-linear maneuver includes: determining whether the detected flying object is the flying object capable of the non-linear maneuver from the detection information updated at the preset cycle and the acquired prior information, using the neural network model; and synthesizing multiple probability values for the determining result output by the neural network model at a preset cycle, comparing the synthesized multiple probability values with a preset threshold value, and re-determining whether the detected flying object is the flying object capable of non-linear maneuver. ### Claim 8 (depends on 7) The method of claim 7, wherein the re-determining whether the detected flying object is the flying object capable of non-linear maneuver contains synthesizing the multiple probability values by assigning a relatively higher weight to the probability value calculated in a latest cycle among the multiple probability values calculated by the neural network model for each preset cycle, and then calculating a synthesis probability value. ### Claim 9 (depends on 8) The method of claim 8, wherein the synthesis probability value is obtained through the following equation: synthesis ⁢ probability ⁢ value = ∑ k = 1 n probability value calculated in k cycle×k cycle weight where k is information identifying which cycle it is, n is a total cycle, and k cycle weight=k/100. ### Claim 10 (depends on 7) The method of claim 7, wherein determining the impact point includes re-determining the impact point of the detected flying object based on the detection information updated at a preset cycle, the prior information, and re-determined whether the detected flying object is the flying object capable of non-linear maneuver. ### Claim 11 A device for predicting an impact point, the device comprising: a communication interface; a memory including an instruction; and a processor configured to perform, by executing the instruction: acquiring detection information related to a flying object detected by a radar linked to the device through the communication interface; acquiring prior information including information about an enemy unit, a friendly unit, and a friendly facility; determining whether the detected flying object is a flying object capable of non-linear maneuver based on the detection information and the prior information, using a neural network model pre-trained using training data so that the neural network model determines whether the flying object is the flying object capable of non-linear maneuver based on detection information and prior information; and determining an impact point based on the detection information, the prior information, and a determining result of the neural network model on whether the detected flying object is the flying object capable of non-linear maneuver, wherein the neural network model includes an input layer set to input detection information and prior information of a flying object, an output layer outputting the determining result, and a model parameter trained to have a correlation between the detection information, the prior information, and the determining result based on a supervised training by using label information specifying whether the flying object is capable of non-linear maneuver. ### Claim 12 (depends on 11) The device of claim 11, wherein acquiring the detection information includes extracting dynamic characteristic information of the detected flying object by replaying an image of the flying object detected by the radar. ### Claim 13 (depends on 12) The device of claim 12, wherein the dynamic characteristic information contains information about a launch point, moving position, moving trajectory, speed, acceleration, and RCS of the detected flying object. ### Claim 14 (depends on 11) The device of claim 11, wherein acquiring the detection information includes extracting dynamic characteristic information of the detected flying object by replaying an image of the flying object detected by the radar, and wherein the neural network model uses the extracted dynamic characteristic information as training data for the detection information that is to be input into the input layer. ### Claim 15 (depends on 11) The device of claim 11, wherein determining whether the detected flying object is the flying object capable of non-linear maneuver includes assigning some unspecified detection information to a preset value and inputting the unspecified detection information into the neural network model, when some of multiple pieces of detection information to be input into the neural network model are unspecified. ### Claim 16 (depends on 11) The device of claim 11, wherein acquiring the detection information includes updating the detection information at a preset cycle. ### Claim 17 (depends on 16) The device of claim 16, wherein determining whether the flying object is the flying object capable of non-linear maneuver includes: determining whether the detected flying object is the flying object capable of the non-linear maneuver from the detection information updated at the preset cycle and the acquired prior information using the neural network model; and synthesizing multiple probability values for the determining result output by the neural network model at a preset cycle, comparing the synthesized multiple probability values with a preset threshold value, and re-determining whether the detected flying object is the flying object capable of non-linear maneuver. ### Claim 18 A non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, comprises an instruction for causing the processor to perform a method comprising: acquiring detection information related to a flying object detected by a radar; acquiring prior information including information about an enemy unit, a friendly unit, and a friendly facility; determining whether the detected flying object is a flying object capable of non-linear maneuver based on the detection information and the prior information, using a neural network model pre-trained using training data so that the neural network model determines whether the flying object is the flying object capable of non-linear maneuver based on detection information and prior information; and determining an impact point based on the detection information, the prior information, and a determining result of the neural network model on whether the detected flying object is the flying object capable of non-linear maneuver, wherein the neural network model includes an input layer set to input detection information and prior information of a flying object, an output layer outputting the determining result, and a model parameter trained to have a correlation between the detection information, the prior information, and the determining result based on a supervised training by using label information specifying whether the flying object is capable of non-linear maneuver.