Dynamic self-adaptive sun backlight interference elimination method based on event star sensor

Abstract

The invention discloses a dynamic self-adaptive sun backlight interference elimination method based on an event star sensor, and mainly solves the problem that a space detector cannot accurately predict a backlight angle and cannot effectively suppress noise interference caused by the backlight angle under the condition of sun backlight. The method comprises the following steps: firstly, establishing a relationship between a threshold value and an event rate under different backlight angles through a support vector regression (SVR) model; secondly, measuring optimal threshold values under different backlight angles by using a receiver operation characteristic (ROC) curve; and finally, based on event flow characteristic analysis in a unit time window, estimating a backlight angle through an SVR model and adaptively adjusting a response threshold. The backlight interference is effectively eliminated in a dynamic environment. In a non-direct backlight environment, the method designed by the invention successfully realizes effective suppression of irrelevant light noise and reduction of event processing cost while guaranteeing the detection precision, and has a relatively high practical value.

Description

Technical Field

[0001] The present invention relates to computer vision, backlight target detection, optimal control and other field technologies. Specifically, it is a space target detection method which uses an event camera to dynamically and adaptively eliminate the interference of solar backlight. Background Art

[0002] Star sensors are space attitude sensors that use star detection as a reference to achieve space navigation, and complete the carrier's attitude measurement and positioning through star map matching. However, the navigation performance of traditional star sensors is significantly limited when dealing with environmental stray light, especially when dealing with backlighting scenarios where the sun is incident at a small angle, which is frequently encountered during the on-orbit operation phase of space-based payloads. Existing technical means are still insufficient to address the interference problem caused by solar stray light on star sensors in backlighting scenarios.

[0003] Traditional solutions focus on adding a sunshade to the hardware and software post-processing algorithms, but they are limited by imaging effects, computational complexity, or environmental adaptability, and are generally ineffective in dealing with dynamic backlit stray light environments. As a new type of visual sensor, event cameras, unlike traditional cameras, only record information when the brightness change of a unit pixel exceeds a threshold, providing a new approach to solving the above problems. Event cameras are not limited by absolute illumination, and perform time-domain brightness differentiation of background light and target signals. They have the advantages of large dynamic range, high sensitivity, high sampling rate, and low bandwidth, and can essentially improve the efficiency and data quality of signal acquisition in backlit environments.

[0004] In the space environment, the relative positions of the detector, space target and the sun are constantly changing. In extreme environments, when the sun enters the detector's field of view, the focused solar photons will cause leakage noise events triggered by extremely high frequencies dominated by junction leakage current. Even when the sun does not appear in the field of view, its light will enter the lens and produce irrelevant stray light noise events. Therefore, when the detector is operating in the backlight area interfered by the strong light of the sun, it is extremely difficult to detect stellar targets. In this context, the dynamic adaptive backlight interference elimination method of the event camera is expected to be an effective means to solve this problem. Summary of the invention

[0005] The present invention provides a dynamic adaptive solar backlight interference elimination method based on an event star sensor, which mainly solves the problem that space detectors cannot accurately predict the backlight angle and effectively suppress the noise interference caused by it when facing solar backlight. This method can adjust the event trigger threshold according to real-time environmental data and optimize the performance of the detector when facing complex lighting changes.

[0006] The technical solution to achieve the purpose of the present invention is: a dynamic adaptive solar backlight interference elimination method based on an event star sensor, the method comprising the following steps:

[0007] Step (1): construct a backlight angle prediction model based on SVR, define the backlight angle and event trigger threshold, and collect event volume data corresponding to different backlight angles as training set samples to train the model;

[0008] Step (2): Use the ROC curve to evaluate the detection performance of the system under different thresholds in the training set of step (1), and find the corresponding optimal event trigger threshold for each backlight angle;

[0009] Step (3): define a fixed time window, read the event stream cyclically, calculate the event rate and read the current hardware threshold parameters;

[0010] Step (4): The event rate data and hardware threshold parameters collected in step (3) are used as input to the pre-trained SVR model to predict the backlight angle collected by the current device. According to the prediction result corresponding to the ROC curve pre-evaluated in step (2), the appropriate threshold parameters are dynamically adjusted and called back.

[0011] Compared with the prior art, the present invention has the following significant advantages: the present invention proposes a dynamic adaptive solar backlight interference elimination method based on an event star sensor, which: (1) adopts an event-based sensor, utilizing the advantages of an event camera that is only sensitive to light intensity changes and is not limited by absolute illumination, and has high dynamic range and high temporal resolution, thereby substantially improving the efficiency and quality of signal acquisition in a backlight environment; (2) adopts an SVR-based backlight angle prediction model, utilizing the excellent generalization ability of SVR, and can maintain a high prediction accuracy when there is large noise in the data; (3) adopts a ROC curve to design a threshold optimization strategy, determines the optimal threshold for each backlight angle condition, and combines the SVR model to dynamically and adaptively adjust the event trigger threshold according to real-time environmental data, thereby optimizing the performance of the detector when facing complex illumination changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 The present invention is a flow chart of a method for dynamically adaptively eliminating solar backlight interference based on an event star sensor.

[0013] Figure 2 This is a diagram showing the relationship between the threshold and the number of events when the backlight angle of the event camera is between 180° and 90°.

[0014] Figure 3 This is a diagram showing the relationship between the threshold and the event quantity when the backlight angle of the event camera is between 160° and 90°.

[0015] Figure 4 This is an example of the ROC curve drawn when the backlight angle of the event camera is 160° to 90°. DETAILED DESCRIPTION

[0016] The present invention will be further described below in conjunction with the accompanying drawings.

[0017] The present application embodiment provides a method for dynamically adaptively eliminating solar backlight interference based on an event star sensor. The method flow is as follows: Figure 1 As shown, the method comprises the following steps:

[0018] Step (1): construct a backlight angle prediction model based on SVR, define the backlight angle and event trigger threshold, and collect event volume data corresponding to different backlight angles as training set samples to train the model;

[0019] Step (2): Use the ROC curve to evaluate the detection performance of the system under different thresholds in the training set of step (1), and find the corresponding optimal event trigger threshold for each backlight angle;

[0020] Step (3): define a fixed time window, read the event stream cyclically, calculate the event rate and read the current hardware threshold parameters;

[0021] Step (4): The event rate data and hardware threshold parameters collected in step (3) are used as input to the pre-trained SVR model to predict the backlight angle collected by the current device. According to the prediction result corresponding to the ROC curve pre-evaluated in step (2), the appropriate threshold parameters are dynamically adjusted and called back.

[0022] Furthermore, step (1) specifically includes the following steps:

[0023] Step (11): Set up a backlighting experiment, define the backlighting angle as the angle between the light source and the imaging direction of the event camera, and set the light source and the star point target position to be fixed. Fix the event camera on a high-precision three-axis turntable, and collect all event data from the backlighting angle of 90° to the backlighting angle of 180°, such as Figure 2 and Figure 3 As shown, the experiment is repeated several times. The event data collected in the above experiment is used as the training set samples.

[0024] Step (12): In the training set samples, define the backlight angle and threshold, as well as the corresponding event volume data under different backlight angles. Construct the feature set x and the target variable y; where x is a two-dimensional array that contains all combinations of backlight angles and thresholds; y is a one-dimensional array that contains the event volume corresponding to each combination in x;

[0025] Step (13): For the feature vector of the sample, where x i ={x i1 ,…x ik} is y i The influencing factor of is k, and the number of influencing factors is k. Then the support vector regression function is:

[0026] f(x)=w T g(x)+bg(x)

[0027] Among them, g(x) is a nonlinear function that maps samples to high-dimensional space, w T is the coefficient of the independent variable function, and b is the bias;

[0028] Step (14): Using the generalized optimal classification surface function that takes into account sampling and maximum classification interval, a dual optimization problem is constructed based on the Lagrange equation and dual problem theory to obtain w described in step (12). T and b value, the SVR function model is obtained as:

[0029]

[0030] Among them, K(x i ,x j ) is the kernel function, a i , a j and is the Lagrange multiplier.

[0031] Furthermore, step (2) specifically includes the following steps:

[0032] Step (21): The ROC curve is represented in a coordinate graph, where the horizontal axis is the false positive rate (FPR) and the vertical axis is the true positive rate (TPR); the true positive rate is defined as: the proportion of all actual positive samples that are correctly identified as positive, and its calculation formula is:

[0033]

[0034] Where TN is the number of true negatives;

[0035] The false positive rate is defined as the proportion of all actual negative samples that are mistakenly identified as positive samples. The calculation formula is:

[0036]

[0037] Where FP is the number of false positives;

[0038] Step (22): For each backlight angle, different TPR and FPR values ​​are collected by changing the threshold, and the ROC image is drawn, such as Figure 4 As shown; each group (FPR, TPR) corresponds to a point in the ROC curve graph, and these points are connected to form a ROC curve for each backlight angle; the optimal threshold is the threshold corresponding to the point closest to the upper left corner of the ROC curve. Specifically, find the threshold that makes The minimum threshold ensures that the true positive rate is maintained while minimizing the false positive rate.

[0039] Furthermore, the specific steps of step (3) are as follows:

[0040] Step (31): Use an event camera to perform target detection and shooting under the condition of sun backlight;

[0041] Step (32): Set the initial threshold and define a fixed time window, read the event stream cyclically, and decode the address event stream and store it in the four-tuple e i =(x i ,y i ,t i ,p i ), where (x i ,y i ) is the event pixel coordinate, t i is the event timestamp, p i is the event polarity;

[0042] Step (33): For the event stream of a fixed time window, calculate the event volume per unit time, i.e., the event rate, and read the current hardware threshold parameters.

[0043] Furthermore, step (4) uses the event rate data and hardware threshold parameters collected in step (3) as input to a pre-trained SVR model to predict the backlight angle collected by the current device, and dynamically adjusts the appropriate threshold parameters and calls back according to the ROC curve pre-evaluated in step (22) corresponding to the prediction result.

[0044] The above is a preferred implementation mode of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea, but the implementation mode of the present invention is not limited to the above content. For those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation mode and application scope. Any changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A dynamic adaptive solar backlight interference elimination method based on event star sensor, characterized in that: The steps include: Step (1): construct a backlight angle prediction model based on SVR, define the backlight angle and event trigger threshold, and collect event volume data corresponding to different backlight angles as training set samples to train the model; Step (2): Use the ROC curve to evaluate the target detection performance at different thresholds at multiple angles in the training set of step (1), and find the corresponding optimal event trigger threshold for each backlight angle; Step (3): define a fixed time window, read the event stream cyclically, calculate the event rate and read the current hardware threshold parameters; Step (4): The event rate data and hardware threshold parameters collected in step (3) are used as input to the pre-trained SVR model to predict the backlight angle collected by the current device. According to the prediction result corresponding to the ROC curve pre-evaluated in step (2), the appropriate threshold parameters are dynamically adjusted and called back.

2. The method for eliminating solar backlight interference based on event star sensor according to claim 1, characterized in that: Step (1) specifically includes the following steps: Step (11): set up a backlighting experiment, define the backlighting angle as the angle between the light source and the imaging direction of the event camera, and set the positions of the light source and the star point target to be fixed; fix the event camera on a high-precision three-axis turntable, and collect all event data from the backlighting angle of 90° to the backlighting angle of 180°, and repeat the experiment multiple times; use the event data collected in the above experiment as training set samples; Step (12): Construct an SVR backlight angle prediction model and train the model based on training set samples.

3. The method for eliminating solar backlight interference based on event star sensor according to claim 1, characterized in that: For the training set data with different thresholds at each backlight angle in step (2), at the only backlight angle, the true positive rate TPR and false positive rate FPR of the data at each threshold are counted, and the TPR and FPR of different thresholds at all backlight angles are counted, and the ROC curves at different angles are plotted; for each backlight angle, the optimal threshold corresponds to the point closest to the upper left corner in the ROC curve, that is, the minimum The threshold corresponding to the point is the optimal threshold under the backlight angle.

4. The method for eliminating solar backlight interference based on event star sensor according to claim 1, characterized in that: Step (3) specifically includes the following steps: Step (31): Use an event camera to perform target detection and shooting under the condition of sun backlight; Step (32): Set the initial threshold and define a fixed time window, read the event stream cyclically, and decode the address event stream and store it in the four-tuple e i =(x i ,y i , t i , p i ), where (x i ,y i ) is the event pixel coordinate, t i is the event timestamp, p i is the event polarity; Step (33): For the event stream of a fixed time window, calculate the event volume per unit time, i.e., the event rate, and read the current hardware threshold parameters.

Images