In order to analyze the reasons for the degradation of star sensor performance and the decrease in attitude measurement accuracy caused by harsh spatial radiation environment, a deep study was conducted on 60Co- γ The influence mechanism of the total ionization dose effect on star pattern recognition of complementary metal oxide semiconductor active pixel sensors (CMOS APS) under radiation environment By building an outfield star observation experimental system and observing the zenith and Orion sky area, 60Co was obtained through experimental processes such as star map data collection, star point extraction, and star map recognition- γ The mechanism of the influence of CMOS APS noise on the average gray level of the star map background and the number of recognized star points after X-ray irradiation is studied, and a recognition algorithm for finding stars obliterated by radiation noise is proposed The quantitative relationships between CMOS APS dark current noise, dark signal non-uniformity noise, and light response non-uniformity noise and star centroid positioning error were established through theoretical derivation The research results indicate that 60Co- γ After X-ray irradiation, the average gray level of the star map background of the star sensor increases, and the number of star point recognition decreases. The noise increase after CMOS APS irradiation leads to an increase in the centroid positioning error of the star point, which affects the attitude positioning accuracy of the star sensor. This research result provides a certain theoretical basis for the design and radiation reinforcement of high-precision star sensors
Star sensors, as an important component of satellite attitude and orbit control systems, play a crucial role in satellite attitude control during launch and in orbit flight The star sensor first takes a star map of the sky, performs star map preprocessing, star map recognition, and attitude calculation, and finally outputs attitude angle data for autonomous navigation of spacecraft such as satellites. The core component of the star sensor is the imaging system Early star sensors used charge coupled device image sensors as star imaging devices. With the increasing demand for low-power and miniaturization of star sensors and the advancement of complementary metal oxide semiconductor (CMOS) technology, star sensors based on CMOS image sensors have become the mainstream product in the current market.
The harsh spatial radiation environment can cause dark current and dark signal non-uniformity (DSNU) in complementary metal oxide semiconductor active pixel sensors (CMOS APS) The degradation of radiation sensitive parameters such as photo response non-uniformity (PRNU) and CMOS image sensor parameters leads to a significant increase in the background noise of the star map collected by the star sensor, which affects the performance of star point centroid positioning, star map recognition, attitude positioning accuracy, and other aspects of the star sensor
A study has found that the star maps collected by space working star sensors after irradiation exhibit performance degradation phenomena such as reduced signal-to-noise ratio and trailing star points, which affects the accuracy of star sensors in orbit attitude positioning At present, satellites are equipped with both star sensors and gyroscopes. When space radiation causes a decrease in the star map recognition ability of star sensors and abnormal attitude positioning, most engineering units adopt the solution of turning off star sensors and relying on gyroscopes for attitude positioning This solution can enable the satellite to continue operating, but there are certain risks because the attitude positioning accuracy of gyroscopes is lower than that of star sensors, and there is uncertainty in satellite attitude control In addition, image processing algorithms can also be used to remove the influence of radiation induced noise on star pattern recognition. However, the establishment of image processing algorithms is based on in-depth analysis of the mechanism of CMOS APS radiation induced noise and research on the mechanism of CMOS APS radiation effect on star pattern recognition of star sensors. Using only conventional noise reduction algorithms may also filter out star signal, or some unfiltered noise points may be mistakenly recognized as star points by star sensors Therefore, there is an urgent need to conduct research on the mechanism of CMOS APS radiation effect on star map recognition
γ After irradiating the CMOS image sensor with X-rays, the dark current increases with the increase of the total ionizing dose, as shown in Figure 2 One reason for the increase in dark current is that ionizing radiation generates a large number of interface states at the Si SiO2 interface, and the density of interface states increases with the increase of accumulated dose The interface state acts as a recombination center to increase the generation rate of dark current. On the other hand, ionizing radiation introduces positive oxide trap charges into the metal dielectric (PMD) layer on the surface of shallow groove isolation and buried photodiodes (PPDs). Although it cannot directly generate dark current as a recombination center, its additional electric field can make the surface depletion region and the depletion region of PPDs come into contact with each other, The electron hole pairs generated by this interface state can drift into the depletion region of PPD, thereby improving the collection efficiency of charge carriers.
DSNU characterizes the non-uniformity of the response of each pixel unit in the image sensor chip under dark field conditions. DSNU increases with the increase of the total ionizing dose, and the ionizing effect produces electron hole pairs. Electrons and holes randomly recombine while migrating, and the random recombination probability of each pixel unit is different, resulting in an increase in DSNU; The defect energy levels generated by the simultaneous radiation effect act as the center of charge carrier generation recombination, which also increases the probability of electron hole recombination and exacerbates the degradation behavior of DSNU
PRNU characterizes the deviation of light response from the average value, expressed as a percentage PRNU increases with the increase of total ionization dose, and the interface state induced by ionization effect reduces the lifetime of photo generated charge carriers, resulting in a decrease in the efficiency of photo generated charge collection, manifested as a degradation of quantum efficiency Due to the uneven distribution of interface trap charges, there are differences in quantum efficiency degradation between pixels, leading to an increase in pixel responsiveness differences and ultimately leading to an increase in PRNU.
(1) Mean background grayscale
Install CMOS image sensors irradiated to 7.5, 10, 20, and 50 krad (Si) into the star sensor testing system, and take photos of the Orion sky region and zenith one by one. Collect 50 frames of images at 95.6, 143.4525.6 ms for each dose point Calculate the average background grayscale of the star map collected by the star sensor at each dose point at different integration times, as listed in Table 1. As the cumulative dose increases, the average background grayscale gradually increases For the same cumulative dose, as the integration time increases, the average background grayscale also gradually increases The main reason for the increase in average background grayscale is the increase in total noise of CMOS image sensors after irradiation.
The noise of CMOS image sensors comes from the reset and readout processes of photodiodes and pixel transistors in their photosensitive units, active amplifiers, row and column selection switches, etc., mainly including dark current noise, DSNU noise, PRNU noise, fixed mode noise, and readout noise Among them, DSNU noise, PRNU noise, and fixed mode noise increase with the increase of cumulative dose The dark current noise increases with both the cumulative dose and the integration time
The dark current noise is equal to the square root of the number of electrons generated by the dark current. The dark current source of CMOS image sensors can be roughly divided into two parts: pixel units and peripheral circuits The working principle of CMOS image sensors determines that the dark current in the peripheral circuit is a fixed value independent of the integration time; As the integration time increases, the output grayscale value of pixel unit dark current will gradually increase Ionizing radiation causes an increase in the differential generation rate of dark current between pixels, leading to an increase in DSNU noise. PRNU noise characterizes the deviation of light response from the average value. Due to the uneven distribution of interface trap charges, there are differences in quantum efficiency degradation between pixels, resulting in an increase in pixel responsiveness differences, ultimately leading to PRNU noise increases.
Fixed mode noise reflects the difference between pixels. At a fixed integration time, fixed mode noise is basically a constant, which mainly comes from two aspects: on the one hand, it is caused by the mismatch of intra pixel transistors or column level transistors during the manufacturing process; On the other hand, it is the dark current within the pixel The fixed mode noise caused by transistor mismatch can be eliminated through correlated double sampling, while the sources of dark current are diverse and the generation mechanisms are different. The fixed image noise caused by dark current cannot be completely eliminated Under no light conditions, fixed mode noise can be characterized by DSNU; Evaluate through PRNU under lighting conditions Read out noise belongs to transient noise, which is caused by random fluctuations in signal levels caused by various noise sources in the circuit channel (column amplifiers, programmable gain amplifiers, and analog-to-digital converters). During the experiment, the circuit channel remains unchanged, so the impact of read out noise can be ignored
(2) Identify the number of star points
Before star map recognition, it is necessary to filter the Smithsonian Astrophysical Observatory star catalog (SAO) based on the limiting magnitude of the star sensor and establish a star diagonal distance lookup table The star to star distance lookup table consists of star to star signs and the cosine value of star to star diagonal distance The cosine value of star diagonal distance can be calculated based on the direction vectors of the two navigation stars A and B that make up the star pair. The cosine value of star diagonal distance for any two stars that match the star sensitive field of view angle can be calculated and sorted from small to large Then preprocess the star map and subdivide the star points for positioning Star map preprocessing first filters the captured star map, and then calculates the threshold used for coarse positioning of stars through corresponding calculations. Through threshold segmentation, connected domain labeling is performed on the stars Subdivision positioning of star points can obtain the accurate position of star points in the star map, such as using the centroid method, which can achieve sub pixel accuracy
Use triangle algorithm for star map recognition The triangle star map recognition method is divided into the following 10 steps
1) Firstly, extract star points from the star map, and then sort them according to their energy, selecting the brightest Sn stars
2) Make the star closest to the optical axis the first target star S1 in the brightest star
3) Outside the radius ring of S1, select the two brightest stars S2 and S3 in the field of view
4) Sort these three stars according to their triangular geometric relationship
5) Calculate the positions of these three stars on the star sensor detector array using the centroid method (XI, YI):
6) Assuming the coordinates of two stars are (x1, y1) and (x2, y2), calculate the direction vectors of these two stars in the star sensor coordinate system based on the calibrated main point positions (x0, y0) and focal length f
7) Calculate the measurement value of the cosine of the diagonal distance between star pairs using the above method
8) Considering the imaging error of the star sensor, there is a small difference between the measured value of the cosine of the star diagonal distance calculated in step 7) and the cosine value of the star diagonal distance in the star pair lookup table. Compare the star diagonal distance lookup table to identify possible navigation star combinations.
9) Find the same stars S1, S2, and S3 from the navigation star combination.
10) If S1, S2, and S3 are unique, recognition is successful; If S1, S2, and S3 are not unique, repeat the above steps and rebuild the triangle for recognition
Analyze the matching results of star map triangle recognition under different cumulative doses, and debug and analyze the star map recognition program to address the issue of a decrease in the number of star point recognition under high cumulative doses. It was found that the number of stars to be identified decreases as the cumulative dose increases In order to re identify unrecognized stars, this article first determines the reason for a certain star point not being recognized based on the imaging principle of the star sensor, the rotation transformation from the celestial coordinate system to the star sensor coordinate system, and the projection transformation relationship from the star sensor coordinate system to the image coordinate system
(3) Accuracy of star point centroid positioning
The accuracy of star sensor attitude measurement mainly depends on the accuracy of star point centroid positioning The centroid positioning error of star points in the star map taken by the star sensor cannot be directly calculated, but can be indirectly calculated through the standard error of star diagonal distance. The standard error of star diagonal distance is 1.42 times the standard error of star centroid positioning Due to the differences in the number of star point recognition under different cumulative doses, when calculating the standard error of star diagonal distance, the other cumulative dose points were selected with the same 5 stars successfully identified with 50 krad (Si) Calculate the standard error of star diagonal distance for 5 stars under different cumulative doses using measured and theoretical star diagonal distance. Based on the relationship between the standard error of star diagonal distance and the standard error of centroid positioning, combined with the field angle and resolution of the star sensor, calculate the standard error of star centroid positioning under different cumulative doses. As the cumulative dose increases, the standard error of star diagonal distance gradually increases, and the standard error of centroid positioning also gradually increases, The accuracy of star point centroid positioning gradually decreases For the same cumulative dose, as the integration time decreases, the standard error of star diagonal distance and centroid positioning gradually increase, and the accuracy of star point centroid positioning gradually decreases According to the analysis, γ Radiation irradiation can cause an increase in dark current noise, DSNU noise, and PRNU noise in CMOS image sensors. The increase in noise after irradiation of the image sensor can lead to an increase in background noise of the star map, thereby affecting the accuracy of star centroid positioning In order to quantitatively analyze the effects of dark current noise, DSNU noise, and PRNU noise of CMOS image sensors on the centroid positioning accuracy of star sensors, functional relationships were established for the effects of each noise on the centroid positioning error
From the above analysis, it can be seen that as the cumulative dose increases, the dark current noise, DSNU noise, and PRNU noise of CMOS image sensors gradually increase, leading to an increase in star centroid positioning error, ultimately affecting the attitude measurement accuracy of the star sensor. Increasing the integration time appropriately can reduce the star centroid positioning error However, if the integration time is too long, the pixels in the star region will approach or reach the saturation zone, and the rate of photon transfer curve change will slow down or no longer change. At this time, increasing the integration time will no longer significantly reduce the centroid positioning error of the star
Co- γ After X-ray irradiation, the dark current noise, DSNU noise, and PRNU noise of CMOS APS increase with the increase of cumulative dose. The increase in total noise of CMOS APS will lead to an increase in the average gray level of the background image captured by the star sensor, with obvious background fluctuations, resulting in an increase in the difficulty of star point recognition and a decrease in the number of star point recognition Based on mechanism analysis, this article introduces an algorithm for finding unrecognized star points. This algorithm can successfully identify star points that have been obscured by background noise. Research has found that the increase in CMOS APS dark current noise, DSNU noise, and PRNU noise leads to the shift of star point centroid position, ultimately affecting the accuracy of star sensor star point centroid positioning This article lays a theoretical foundation for star sensor design units to improve the success rate of star sensor recognition and ensure the safe and reliable operation of satellites in orbit.
Send us a message,we will answer your email shortly!
TY-Space Technology (Beijing) Ltd. is professional focusing on advanced attitude optical sensors, especially Star Trackers for space industry.
