Methods and systems for non-cooperative automatic security screening in crowded areas
The present invention is a multi-threat detection method and system for high-traffic areas, such as airports, other mass-transit hubs, malls, symposiums, and the like. The detection methodology comprises a set of detection blocks, each performing different functions, and taking the information resulting from each block and entering it into one or more detection algorithms. The blocks comprise microwave detectors, co-polarization versus cross-polarization processors, magnetometers, and cameras (optional). The values calculated by the blocks are entered into machine learning algorithms which detect a threat if a threshold value is met. The detection occurs at a distance and the alarm may be silent, such that authorities are able to address threats without the target being aware. The alarm is also able to differentiate among varying threats based on the values calculated. No shadowing problems exist, and thus several targets may be monitored simultaneously without slowing down the speed of processing.
1. A method for detecting one or more threats on one or more targets passing through a security area, comprising:
calculating a dielectric permittivity value of one or more objects on said one or more targets via one or more microwave signals, said dielectric permittivity value corresponding to a variation in a microwave reflected or transmitted through said one or more objects,
determining a magnetic strength effect value of said one or more objects on said one or more targets via one or more magnetometers,
entering said calculated dielectric permittivity value and said magnetic strength effect value into at least one machine learning algorithm, and
creating an alarm after one screening stage if said one or more algorithms indicate a threshold value, said threshold value indicating a threat.
2. The method of claim 1 , further comprising the step of:
processing co-polarization versus cross-polarization values of said one or more microwave signals, thus forming a polarization ratio value for said one or more objects, and entering said polarization ratio value into said one or more machine learning algorithms along with said dielectric permittivity value and said magnetic strength effect value.
3. The method of claim 2 , wherein said dielectric permittivity value and said polarization ratio value are collected at a speed of at least 15 frames per second (fps).
4. The method of claim 1 , further comprising Principal Component Analysis (PCA) pre-processing.
5. The method of claim 1 , comprising at least two algorithms, wherein either algorithm may indicate a threat without a need for both algorithms to indicate said threat.
6. The method of claim 1 , wherein at least one algorithm is a Mahalanobis distance algorithm.
7. The method of claim 1 , wherein at least one algorithm is a support vector machine (SVM) algorithm.
8. The method of claim 1 , further comprising performing a noise cutoff using frequency filtering.
9. The method of claim 1 , wherein at least two magnetometers are employed to determine said magnetic strength effect value.
10. The method of claim 9 , wherein results from a first magnetometer are compared with results from a second magnetometer.
11. The method of claim 9 , wherein the at least two magnetometers form a portal up to 2.5 meters wide.
12. The method of claim 10 , further comprising a camera, said camera taking a photo, said photo being synchronous with said compared results between said magnetometers.
13. The method of claim 1 , further comprising performing a facial recognition of a target, wherein said facial recognition is performed via one or more video cameras.
14. The method of claim 1 , wherein at least 2 video cameras are employed to construct a 3D video image of the one or more objects, said 3D video image being complementary to a 3D microwave image of said objects, said at least 2 video cameras being synchronized in time with said one or more microwave signals.
15. The method of claim 1 , wherein the alarm differentiates among a target without objects, a target with a metal, and a target with a dielectric.
16. The method of claim 1 , wherein the alarm differentiates among benign metal objects and threatening metal objects.
17. The method of claim 1 , further comprising displaying results via a graph, said graph differentiating an algorithmic output value via color or shape or both color and shape.
18. The method of claim 1 , wherein the alarm is a silent alarm.
19. The method of claim 1 , said method being capable of detecting metallic objects being shadowed by one or more other targets within the area.
20. The method of claim 1 , said method being capable of detecting metallic objects inside said target's body.
21. A method for detecting one or more threats on one or more targets passing through a security area, comprising:
calculating a dielectric permittivity value of one or more objects on said one or more targets via one or more microwave signals, said dielectric permittivity value being collected at a speed of at least 15 frames per second,
determining a magnetic strength effect value of said one or more objects on said one or more targets via one or more magnetometers,
entering said calculated dielectric permittivity value and said magnetic strength effect value into at least one machine learning algorithm, and
creating an alarm if said one or more algorithms indicate a threshold value, said threshold value indicating a threat.