Space debris mainly refers to all man-made objects except for normal spacecraft, including various satellites that have completed their missions, rocket bodies, waste in the process of performing space missions, and debris generated by the collision of space objects. These debris pose a great threat to the safety of spacecraft in orbit. With the continuous development of science and technology, space activities are becoming more and more frequent, which has led to an increasing number of space debris, posing a serious threat to the safety of spacecraft in orbit. Therefore, in order to reduce the impact of space debris on spacecraft, it is particularly important to carry out orbital monitoring and prediction of debris. Optical cameras play an important role in the field of space debris detection with their advantages of low energy consumption, high resolution and wide field of view coverage. Traditional initial orbit determination algorithms such as the Laplace method and the Gauss method are difficult to achieve the initial orbit determination of targets under short arc conditions due to the large error influence under short arc conditions. Some new methods based on dense data require optical cameras to continuously shoot targets. On the one hand, a large amount of data increases the burden on the system. On the other hand, continuous exposure of high-pixel optical cameras will cause overheating problems and affect the stability of the camera. Therefore, this paper proposes to use spatial filtering method to measure the angular velocity of the target, build the corresponding fitness function based on sparse angle data, and use genetic algorithm to optimize the initial orbit determination of the target.This paper simulates and verifies this method through simulation imaging. First, the orbital parameter distribution of the current space debris is statistically analyzed (Fig.2-Fig.3). Then a simulation observation model is established in this paper (Fig.4). The target is detected by the observation equipment, and a part is received by the high-precision and low-exposure frequency imaging detector, which records the images of the target at three moments: when it enters the field of view, in the field of view, and before it leaves the field of view (Fig.6). The angle information of the target at these three moments is obtained through astronomical positioning, and the other part is filtered through a filter (Fig.1). The high-frequency simulation image is convolved with the sine filter to obtain the time domain brightness signal of the target, and is focused on the photometric sensor through the lens. At this time, the photometric sensor only receives the brightness information of the target (Fig.7). By optimizing the fitness function using a genetic algorithm, the orbital parameter error of the target is obtained (Tab.6)By simulating four low-orbit targets, the target angular velocity error is within 5% through the spatial filtering method (Tab.3). The fitness function is optimized by genetic algorithm, and the distribution of the fitness function in the solution space is obtained (Fig.8). The results show that the relative errors of the distance between the detector and the target are 1.94%, 0.00%, 0.78% and 1.57%, respectively, and the distance measurement error is less than 40 km; the relative errors of the distance change rate are 6.25%, 6.42%, 8.01%, 4.12% (Tab.5), respectively, and the measurement error of the distance change rate is less than 200 m/s, which shows that this method has good ranging accuracy. By solving the target position and velocity information, the target orbit parameter error (Tab.6) is obtained, the semi-major axis error is less than 110 km, the eccentricity error is less than 0.05, and the orbit inclination error is less than 0.8°. Compare the results with the Laplace method and the Gauss method (Tab.7). This method shows good initial orbit determination accuracy for low-orbit targets.This paper proposes a method of using spatial filtering velocimetry to measure angular velocity and combine sparse angle numbers to determine the initial orbit of the target. This method uses continuous time-domain brightness signals and a small amount of angle information. Compared with the traditional method of continuously shooting the target with an optical camera to obtain angle information, this method effectively reduces the amount of data for initial orbit determination and reduces the workload of the optical camera. Secondly, by statistically analyzing the orbital parameter distribution of low-orbit space debris, the results show that the semi-axis length of most space debris is between 6800 km and 8000 km, and the eccentricity is concentrated within 0.05. Based on this, the evaluation of eccentricity is added to the fitness function, and the simulation of four targets is verified. The results show that the semi-major axis error of the four targets is less than 110 km, the eccentricity error is less than 0.05, and the orbit inclination error is less than 0.8°, which shows that the initial orbit determination accuracy of low eccentricity orbit targets is better. The reason is that the evaluation of eccentricity is added to the fitness function, which enhances the convergence of the fitness function and suppresses the ambiguity of the solution to a certain extent. However, this method is mainly applicable to low-orbit, low-eccentricity targets. In the future, this method will be further improved to expand the scope of applicable targets and provide more extensive support for space debris orbit determination and target situation awareness.