Target vehicle path determination method based on encrypted feature vector

Abstract

The invention relates to the technical field of path determination, in particular to a target vehicle path determination method based on encrypted feature vectors. The method comprises the following steps: determining whether a target vehicle is an abnormal vehicle according to videos that the target vehicle enters and leaves a target position area, if so, acquiring vehicle feature data corresponding to the target vehicle, and encrypting the vehicle feature data corresponding to the target vehicle to obtain an encrypted feature vector corresponding to the target vehicle; sending the encrypted feature vector corresponding to the target vehicle to a video processing platform, so that the video processing platform matches the encrypted feature vector corresponding to the target vehicle with encrypted feature vectors in the initial encrypted feature vector set to obtain a target encrypted feature vector set; and determining a driving path of the target vehicle according to the time and the position corresponding to the encrypted feature vector in the target encrypted feature vector set. According to the invention, the safety and accuracy of the abnormal vehicle path determination process can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle path determination, and in particular to a target vehicle path determination method based on encrypted feature vectors. Background Art

[0002] The ETC system (Electronic Toll Collection system) facilitates the rapid passage of vehicles. However, there may be some problems. For example, a vehicle may follow the preceding vehicle into the ETC lane, that is, the vehicle passes through the ETC barrier within the barrier lifting time triggered by the ETC device of the preceding vehicle, and does not use its own ETC device to lift the barrier. In order to determine the driving path of the above-mentioned abnormal vehicle (that is, the vehicle following the preceding vehicle into or out of the ETC lane) and reasonably determine the toll amount of the abnormal vehicle, the abnormal vehicle judgment system needs to send the characteristics of the above-mentioned abnormal vehicle to the video processing platform of the vehicle on the highway so that the video processing platform can capture the location information of the above-mentioned abnormal vehicle. However, if the characteristics of the abnormal vehicle are directly sent to the video processing platform for direct feature matching, it is not conducive to protecting the feature information of the above-mentioned abnormal vehicle and there is an insecurity issue. If only the characteristics that can uniquely characterize the abnormal vehicle (such as the license plate number) are encrypted and sent to the video processing platform for feature matching, since the license plate number of the abnormal vehicle may also change during driving, it may be impossible to match successfully, resulting in the inability to accurately obtain the path of the abnormal vehicle. How to improve the security and accuracy of the abnormal vehicle path determination process is an urgent problem to be solved. Summary of the Invention

[0003] The present invention aims to provide a target vehicle path determination method based on encrypted feature vectors to improve the security and accuracy of the abnormal vehicle path determination process.

[0004] According to the present invention, a method for determining a target vehicle path based on an encrypted feature vector is provided, the method comprising the following steps: S100, determining whether the target vehicle is an abnormal vehicle based on the video of the target vehicle entering and exiting the target location area, and if so, proceeding to S200.

[0005] S200, obtaining vehicle feature data corresponding to the target vehicle, and encrypting the vehicle feature data corresponding to the target vehicle to obtain an encrypted feature vector corresponding to the target vehicle; the vehicle feature data includes a number of vehicle static feature data and a number of vehicle driving behavior feature data; wherein, when the vehicle feature data corresponding to the target vehicle is numerical feature data, Gaussian noise encryption processing is adopted, and the standard deviation corresponding to the Gaussian noise added during the Gaussian noise encryption processing is positively correlated with the sensitive discrimination coefficient of the corresponding vehicle feature, the sensitive discrimination coefficient of the vehicle feature is obtained according to the sensitivity and discrimination corresponding to the vehicle feature, the sensitive discrimination coefficient of the vehicle feature is positively correlated with the sensitivity corresponding to the vehicle feature, and the sensitive discrimination coefficient of the vehicle feature is negatively correlated with the discrimination corresponding to the vehicle feature.

[0006] S300, sending the encrypted feature vector corresponding to the target vehicle to the video processing platform, so that the video processing platform matches the encrypted feature vector corresponding to the target vehicle with the encrypted feature vector in the initial encrypted feature vector set to obtain a target encrypted feature vector set; the target encrypted feature vector set includes the encrypted feature vector in the initial encrypted feature vector set that matches the encrypted feature vector corresponding to the target vehicle.

[0007] S400, determining a driving path of a target vehicle according to a time and a position corresponding to an encrypted feature vector in a target encrypted feature vector set.

[0008] Compared with the prior art, the present invention has at least the following beneficial effects: After determining that a target vehicle is an abnormal vehicle, the present invention obtains multiple feature data of the target vehicle, the multiple feature data including static feature data and driving behavior feature data of the vehicle, and encrypts the feature data of the target vehicle and transmits it to a video processing platform, so that the video processing platform compares feature vectors according to the encrypted feature vectors, and thus obtains the position where the target vehicle has passed; the object of the comparison process of the present invention is the encrypted feature vector, so the comparison process will not leak direct feature information of the target vehicle, and the security is higher; and the present invention uses multiple feature data of the target vehicle for comparison. Compared with the method of performing feature comparison using only one feature that can uniquely characterize the target vehicle but may change, the present invention can perform feature comparison based on multiple features of the target vehicle, and the comparison result will be more accurate; thus, the present invention can improve the security and accuracy of the abnormal vehicle path determination process.

[0009] Moreover, the present invention adopts Gaussian noise encryption processing for numerical feature data, and the standard deviation of the Gaussian noise added during the Gaussian noise encryption processing is positively correlated with the sensitivity discrimination coefficient of the corresponding vehicle feature. Therefore, the noise added to the vehicle feature with higher sensitivity and lower discrimination is larger, and the noise added to the vehicle feature with lower sensitivity and higher discrimination is smaller. This can achieve the purpose of preventing the vehicle feature with higher sensitivity and lower discrimination from being leaked and preventing the vehicle feature with lower sensitivity and higher discrimination from being distorted, further improving the security and accuracy of the abnormal vehicle path determination process. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 A flow chart of a method for determining a target vehicle path based on an encrypted feature vector provided by an embodiment of the present invention; Figure 2 A flowchart of the matching process provided by an embodiment of the present invention; Figure 3 A flowchart for determining the driving path of a target vehicle provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0012] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0013] According to this embodiment, Figure 1 As shown, a method for determining a target vehicle path based on an encrypted feature vector is provided, the method comprising the following steps: S100, determining whether the target vehicle is an abnormal vehicle based on the video of the target vehicle entering and exiting the target location area, and if so, proceeding to S200.

[0014] As a specific implementation method, the target location area is the ETC transaction sensing area, and the abnormal vehicle is a vehicle that has abnormal behavior of passing through the ETC barrier; optionally, if the target vehicle passes through the target location area within the barrier lifting time triggered by the ETC device of the preceding vehicle and the target vehicle does not trigger the barrier lifting, the target vehicle is determined to be an abnormal vehicle.

[0015] In this embodiment, if the target vehicle is not an abnormal vehicle, the process continues to determine whether the vehicle in the subsequent video is an abnormal vehicle, and steps S200 to S400 are not executed.

[0016] S200, obtaining vehicle feature data corresponding to the target vehicle, and encrypting the vehicle feature data corresponding to the target vehicle to obtain an encrypted feature vector corresponding to the target vehicle; the vehicle feature data includes a number of vehicle static feature data and a number of vehicle driving behavior feature data; wherein, when the vehicle feature data corresponding to the target vehicle is numerical feature data, Gaussian noise encryption processing is adopted, and the standard deviation corresponding to the Gaussian noise added during the Gaussian noise encryption processing is positively correlated with the sensitive discrimination coefficient of the corresponding vehicle feature, the sensitive discrimination coefficient of the vehicle feature is obtained according to the sensitivity and discrimination corresponding to the vehicle feature, the sensitive discrimination coefficient of the vehicle feature is positively correlated with the sensitivity corresponding to the vehicle feature, and the sensitive discrimination coefficient of the vehicle feature is negatively correlated with the discrimination corresponding to the vehicle feature.

[0017] In this embodiment, the categorized vehicle features are qualitative features, which are used to describe vehicle attributes and whose differences cannot be measured by relative numerical values, such as vehicle type and vehicle color. Numerical vehicle features are features that can be measured by relative numerical values, such as vehicle height, vehicle width, and vehicle length.

[0018] In this embodiment, the values ​​of the vehicle's static feature data are fixed and do not change during high-speed driving. For example, vehicle static features include vehicle type, vehicle color, vehicle height, vehicle width, and vehicle length. Vehicle static feature data includes vehicle type data, vehicle color data, vehicle height data, vehicle width data, and vehicle length data. Vehicle type data and vehicle color data are categorized feature data, while vehicle height, vehicle width, and vehicle length are numerical feature data.

[0019] As a specific implementation, each numerical feature data corresponding to the target vehicle is greater than or equal to 0 and less than or equal to 1. As a specific implementation, the vehicle type data of the target vehicle is 0 (vehicle type data 0 indicates that the vehicle type is a sedan; vehicle type data 1 indicates that the vehicle type is a sports utility vehicle; vehicle type data 2 indicates that the vehicle type is a multi-purpose vehicle; vehicle type data 3 indicates that the vehicle type is a sports car, etc.), the vehicle color data is 1 (vehicle color data 0 indicates that the vehicle color is white, vehicle color data 1 indicates that the vehicle color is black, vehicle color data 2 indicates that the vehicle color is silver, vehicle color data 3 indicates that the vehicle color is gray, vehicle color data 4 indicates that the vehicle color is gray). Indicates that the vehicle color is red, vehicle color data 5 indicates that the vehicle color is blue, and vehicle color data 6 indicates that the vehicle color is yellow), body height data is 0.93 (body height data 0.93 indicates that the body height is 0.93×H, and H is the preset maximum body height), body width data is 0.75 (body width data 0.75 indicates that the body width is 0.75×W, and W is the preset maximum body width), and body length data is 0.8 (body length data 0.8 indicates that the body length is 0.8×L, and L is the preset maximum body length).

[0020] The vehicle driving behavior characteristic data can be used to characterize the driving habits of the driver driving the vehicle. Optionally, the driving behavior characteristic data of the target vehicle is the driving behavior characteristic data of the target vehicle when it was previously driving on the highway. Different drivers have different driving behaviors, and their driving behaviors are unique. Moreover, over a period of time, the driver's driving style is relatively stable, without frequent and drastic changes, and the driving behavior is relatively stable. Therefore, the target vehicle's driving behavior characteristic data can be used for vehicle matching. Vehicle driving behavior characteristics include following distance on highways, speed change pattern, average speed, and lane change frequency type. Among them, following distance and average speed on highways are numerical features, while speed change pattern and lane change frequency type are categorical features. For example, the following distance of the target vehicle is 0.7 (the following distance of the target vehicle is 0.7, which means that the following distance of the target vehicle is 0.7×D, where D is the preset maximum following distance), the average speed is 0.92 (the average speed of the target vehicle is 0.92, which means that the average speed of the target vehicle is 0.92×V, where V is the preset maximum average speed), the speed change pattern is 1 (a speed change pattern of 0 indicates that the speed fluctuation is small, and a speed change pattern of 1 indicates that the speed fluctuation is large), and the lane change frequency type is 0 (a lane change frequency type of 0 indicates frequent lane changes, and a lane change frequency type of 1 indicates infrequent lane changes).

[0021] In this embodiment, the encrypted feature vector corresponding to the target vehicle includes a deterministic feature subvector and a random feature subvector. The deterministic feature subvector is composed of the deterministic encryption processing result corresponding to the categorical feature data corresponding to the target vehicle, and the random feature subvector is composed of the noise encryption processing result corresponding to the numerical feature data corresponding to the target vehicle. Optionally, the deterministic encryption processing result is a hash encryption processing result.

[0022] In this embodiment, the vehicle features to be acquired are preset, and whether any vehicle feature is a categorical feature or a numerical feature is also preset. The sensitivity discrimination coefficient, discrimination degree, and sensitivity of any numerical feature are also preset. The greater the sensitivity of a vehicle feature, the more private the vehicle feature is, and the more the corresponding data needs to be prevented from being leaked. The greater the discrimination degree of a vehicle feature, the better the data of the vehicle feature can distinguish different vehicles. As a preferred embodiment, the sensitivity discrimination coefficient of the jth numerical feature corresponding to the target vehicle is a j , a j =k1×b j +k2×(1-c j ), b j is the sensitivity of the j-th numerical feature corresponding to the target vehicle, c j is the discrimination of the jth numerical feature corresponding to the target vehicle, k1 and k2 are the preset sensitivity weight and discrimination weight respectively, k1 and k2 are both greater than 0 and less than 1, k1+k2=1, b j and c j are both greater than or equal to 0 and less than or equal to 1, the value range of j is 1 to m, and m is the number of numerical features. Based on this preferred embodiment, the sensitivity discrimination coefficient of any numerical feature can be obtained.

[0023] As a preferred embodiment, the standard deviation of the Gaussian noise added when Gaussian noise is used to encrypt the j-th numerical feature data corresponding to the target vehicle is σ j , σ j =ε / (p×(2-a j )), ε is the preset deviation threshold, p is the preset adjustment coefficient, p≥2, 0≤a j≤1. Based on this preferred specific embodiment, when each numerical feature data corresponding to the target vehicle is greater than or equal to 0 and less than or equal to 1, the standard deviation corresponding to the added Gaussian noise increases as the sensitive discrimination coefficient increases, and decreases as the sensitive discrimination coefficient decreases, so that the noise value added to the data of the numerical feature with a larger sensitive discrimination coefficient is more likely to be greater than the noise value added to the data of the numerical feature with a smaller sensitive discrimination coefficient. This can simultaneously prevent the vehicle features with a larger sensitive discrimination coefficient from being leaked and prevent the vehicle features with a smaller sensitive discrimination coefficient from being significantly distorted, further improving the security and accuracy of the abnormal vehicle path determination process.

[0024] This embodiment uses Gaussian noise encryption processing for numerical feature data, and the standard deviation of the Gaussian noise added during the Gaussian noise encryption processing is positively correlated with the sensitivity discrimination coefficient of the corresponding vehicle feature. Therefore, the noise added to the vehicle feature with higher sensitivity and lower discrimination is larger, and the noise added to the vehicle feature with lower sensitivity and higher discrimination is smaller. This can achieve the purpose of preventing the vehicle feature with higher sensitivity and lower discrimination from being leaked and preventing the vehicle feature with lower sensitivity and higher discrimination from being distorted, further improving the security and accuracy of the abnormal vehicle path determination process.

[0025] S300, sending the encrypted feature vector corresponding to the target vehicle to the video processing platform, so that the video processing platform matches the encrypted feature vector corresponding to the target vehicle with the encrypted feature vector in the initial encrypted feature vector set to obtain a target encrypted feature vector set; the target encrypted feature vector set includes the encrypted feature vector in the initial encrypted feature vector set that matches the encrypted feature vector corresponding to the target vehicle.

[0026] In this embodiment, the initial encrypted feature vector set is obtained by encrypting the vehicle feature data corresponding to vehicles traveling on the target road network during a target time period. Each encrypted feature vector in the initial encrypted feature vector corresponds to a monitoring point, a time, and a location. For example, if a particular encrypted feature vector in the initial encrypted feature vector is obtained by encrypting the vehicle feature data corresponding to a vehicle passing through a first monitoring point, then the first monitoring point is the monitoring point corresponding to the encrypted feature vector, the time when the vehicle passed through the first monitoring point is the time corresponding to the encrypted feature vector, and the location of the first monitoring point is the location corresponding to the encrypted feature vector. The target time period is the predicted time period during which the target vehicle will travel on the highway. The duration of the time period can be an empirical value, such as 2 hours, 4 hours, 8 hours, 12 hours, or 24 hours. For example, if the target vehicle is determined to be an abnormal vehicle at the entrance of the highway, the target time period is determined to be several hours after the time the target vehicle enters the highway. If the target vehicle is determined to be an abnormal vehicle at the exit of the highway, the target time period is determined to be several hours before the time the target vehicle exits the highway. The target road network is the road network in the area where the target vehicle is predicted to travel on the highway, which can be an empirical value, such as the entire highway network; the process of encrypting the vehicle feature data corresponding to the vehicles traveling on the road network is the same as the above-mentioned process of encrypting the vehicle feature data corresponding to the target vehicle, and will not be repeated here.

[0027] like Figure 2 As shown, the matching process includes: S310 , comparing a deterministic feature subvector in a designated encrypted feature vector with a deterministic feature subvector in an encrypted feature vector corresponding to a target vehicle; the designated encrypted feature vector is any encrypted feature vector in an initial encrypted feature vector set.

[0028] In this embodiment, the deterministic feature sub-vector in the designated encrypted feature vector and the elements in the same position in the deterministic feature sub-vector in the encrypted feature vector corresponding to the target vehicle represent the same deterministic feature.

[0029] S320: If the comparison result is the same, proceed to S330; otherwise, determine that the designated encrypted feature vector does not match the encrypted feature vector corresponding to the target vehicle.

[0030] In this embodiment, if all elements in the same position of the deterministic feature subvector in the specified encrypted feature vector and the deterministic feature subvector in the encrypted feature vector corresponding to the target vehicle are the same, the comparison result is determined to be the same; otherwise, the comparison result is determined to be different.

[0031] S330 , obtaining a similarity between a random feature subvector in a designated encrypted feature vector and a random feature subvector in an encrypted feature vector corresponding to a target vehicle.

[0032] As a preferred embodiment, the similarity between the random feature subvector in the encrypted feature vector and the random feature subvector in the encrypted feature vector corresponding to the target vehicle is specified as r, r=1-∑ n i=1 (w i ×d i ), d i is the difference between the i-th element in the random feature subvector in the normalized specified encrypted feature vector and the i-th element in the random feature subvector in the encrypted feature vector corresponding to the target vehicle, w i is the weight of the numerical feature corresponding to the i-th element in the random feature sub-vector, w i The sensitivity coefficient of the numerical feature corresponding to the i-th element in the random feature subvector is negatively correlated, ∑ n i=1 w i =1,w i is greater than 0 and less than 1, and the value range of i is 1 to n, where n is the number of elements included in the random feature sub-vector. i The sensitive discrimination coefficient of the numerical feature corresponding to the i-th element in the random feature subvector is negatively correlated, which can improve the matching accuracy.

[0033] S340: If the similarity is greater than or equal to a preset similarity threshold, it is determined that the designated encrypted feature vector matches the encrypted feature vector corresponding to the target vehicle; otherwise, it is determined that the designated encrypted feature vector does not match the encrypted feature vector corresponding to the target vehicle.

[0034] In this embodiment, the preset similarity threshold is an empirical value.

[0035] Based on S310-S340, the encrypted feature vector in the initial encrypted feature vector set that matches the encrypted feature vector corresponding to the target vehicle can be determined, the encrypted feature vector in the initial encrypted feature vector set that matches the encrypted feature vector corresponding to the target vehicle is also the encrypted feature vector corresponding to the target vehicle, and the time and position corresponding to the encrypted feature vector in the initial encrypted feature vector set that matches the encrypted feature vector corresponding to the target vehicle are the relevant data generated by the target vehicle at different positions when driving on the highway.

[0036] S400, determining a driving path of a target vehicle according to a time and a position corresponding to an encrypted feature vector in a target encrypted feature vector set.

[0037] As an optional specific implementation method, Figure 3 As shown, S400 includes: S410 , arranging positions corresponding to the encrypted feature vectors in the target encrypted feature vector set according to corresponding chronological order.

[0038] S420: Generate a subpath between any two adjacent locations according to the road network.

[0039] It should be understood that a sub-path is part of a road network.

[0040] S430: Determine a driving path of the target vehicle according to the sub-path.

[0041] As an optional specific implementation, the driving path of the target vehicle is formed by connecting the sub-paths in sequence.

[0042] After determining that the target vehicle is an abnormal vehicle, this embodiment obtains multiple feature data of the target vehicle, which include static feature data and driving behavior feature data of the vehicle, and encrypts the feature data of the target vehicle and transmits it to the video processing platform, so that the video processing platform compares the feature vectors according to the encrypted feature vectors, and then obtains the position where the target vehicle has passed; the object of the comparison process of this embodiment is the encrypted feature vector, so the comparison process will not leak the direct feature information of the target vehicle, and the security is higher; and this embodiment uses multiple feature data of the target vehicle for comparison. Compared with the method of performing feature comparison using only one feature that can uniquely characterize the target vehicle but may change, this embodiment can perform feature comparison based on multiple features of the target vehicle, and the comparison result will be more accurate; thus, this embodiment can improve the security and accuracy of the abnormal vehicle path determination process.

[0043] Although some specific embodiments of the present invention have been described in detail by way of example, it should be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present invention. It should also be understood by those skilled in the art that various modifications may be made to the embodiments without departing from the scope and spirit of the present invention. The scope of the present invention is defined by the appended claims.

Claims

1. A method for determining a target vehicle path based on an encrypted feature vector, characterized in that: The method comprises the following steps: S100, determining whether the target vehicle is an abnormal vehicle based on the video of the target vehicle entering and exiting the target location area, and if so, proceeding to S200; S200, obtaining vehicle feature data corresponding to a target vehicle, and encrypting the vehicle feature data corresponding to the target vehicle to obtain an encrypted feature vector corresponding to the target vehicle; the vehicle feature data includes a plurality of vehicle static feature data and a plurality of vehicle driving behavior feature data; wherein, when the vehicle feature data corresponding to the target vehicle is numerical feature data, Gaussian noise encryption is performed, and the standard deviation of the Gaussian noise added during the Gaussian noise encryption is positively correlated with the sensitivity discrimination coefficient of the corresponding vehicle feature, the sensitivity discrimination coefficient of the vehicle feature is obtained based on the sensitivity and discrimination corresponding to the vehicle feature, the sensitivity discrimination coefficient of the vehicle feature is positively correlated with the sensitivity corresponding to the vehicle feature, and the sensitivity discrimination coefficient of the vehicle feature is negatively correlated with the discrimination corresponding to the vehicle feature; S300, sending the encrypted feature vector corresponding to the target vehicle to the video processing platform, so that the video processing platform matches the encrypted feature vector corresponding to the target vehicle with the encrypted feature vectors in the initial encrypted feature vector set to obtain a target encrypted feature vector set; the target encrypted feature vector set includes the encrypted feature vectors in the initial encrypted feature vector set that match the encrypted feature vector corresponding to the target vehicle; S400, determining a driving path of a target vehicle according to a time and a position corresponding to an encrypted feature vector in a target encrypted feature vector set.

2. The target vehicle path determination method based on encrypted feature vector according to claim 1, characterized in that: The encrypted feature vector corresponding to the target vehicle includes a deterministic feature sub-vector and a random feature sub-vector. The deterministic feature sub-vector is composed of the deterministic encryption processing result corresponding to the classified feature data corresponding to the target vehicle, and the random feature sub-vector is composed of the noise encryption processing result corresponding to the numerical feature data corresponding to the target vehicle.

3. The target vehicle path determination method based on encrypted feature vector according to claim 2, characterized in that: The matching process includes: S310, comparing a deterministic feature subvector in a designated encrypted feature vector with a deterministic feature subvector in an encrypted feature vector corresponding to a target vehicle; the designated encrypted feature vector is any encrypted feature vector in the initial encrypted feature vector set; S320, if the comparison result is the same, proceed to S330; otherwise, determine that the specified encrypted feature vector does not match the encrypted feature vector corresponding to the target vehicle; S330, obtaining a similarity between a random feature subvector in a specified encrypted feature vector and a random feature subvector in an encrypted feature vector corresponding to a target vehicle; S340: If the similarity is greater than or equal to a preset similarity threshold, it is determined that the designated encrypted feature vector matches the encrypted feature vector corresponding to the target vehicle; otherwise, it is determined that the designated encrypted feature vector does not match the encrypted feature vector corresponding to the target vehicle.

4. The target vehicle path determination method based on encrypted feature vector according to claim 3, characterized in that: The similarity between the random feature subvector in the encrypted feature vector and the random feature subvector in the encrypted feature vector corresponding to the target vehicle is r, r=1-∑ n i=1 (w i ×d i ), d i is the difference between the i-th element in the random feature subvector in the normalized specified encrypted feature vector and the i-th element in the random feature subvector in the encrypted feature vector corresponding to the target vehicle, w i is the weight of the numerical feature corresponding to the i-th element in the random feature sub-vector, w i Negatively correlated with the sensitivity coefficient of the numerical feature corresponding to the i-th element in the random feature subvector, ∑ n i=1 w i =1,w i is greater than 0 and less than 1, and the value range of i is 1 to n, where n is the number of elements included in the random feature sub-vector.

5. The target vehicle path determination method based on encrypted feature vector according to claim 1, characterized in that: The sensitivity coefficient of the j-th numerical feature corresponding to the target vehicle is a j , a j =k1×b j +k2×(1-c j ), b j is the sensitivity of the j-th numerical feature corresponding to the target vehicle, c j is the discrimination of the jth numerical feature corresponding to the target vehicle, k1 and k2 are the preset sensitivity weight and discrimination weight respectively, k1 and k2 are both greater than 0 and less than 1, k1+k2=1, b j and c j are both greater than or equal to 0 and less than or equal to 1, the value range of j is 1 to m, and m is the number of numerical features.

6. The target vehicle path determination method based on encrypted feature vector according to claim 1, characterized in that: S100 includes: if the target vehicle passes through the target position area within the barrier lifting time triggered by the ETC device of the preceding vehicle, and the target vehicle does not trigger the barrier lifting, then determining that the target vehicle is an abnormal vehicle.

7. The target vehicle path determination method based on encrypted feature vector according to claim 1, characterized in that: Each numerical feature data corresponding to the target vehicle is greater than or equal to 0 and less than or equal to 1.

8. The target vehicle path determination method based on encrypted feature vector according to claim 1, characterized in that: S400 includes: S410, arranging positions corresponding to the encrypted feature vectors in the target encrypted feature vector set according to corresponding time sequence; S420, generating a subpath between any two adjacent locations according to the target road network; S430: Determine a driving path of the target vehicle according to the sub-path.

9. The method for determining a target vehicle path based on an encrypted feature vector according to claim 1, wherein: The initial encrypted feature vector set is obtained by encrypting vehicle feature data corresponding to vehicles traveling on the target road network within a target time period.

10. The target vehicle path determination method based on encrypted feature vector according to claim 2, characterized in that: The deterministic encryption processing result is a hash encryption processing result.

Images