CN-122023761-A - Target detection method and system based on pulse vision

CN122023761ACN 122023761 ACN122023761 ACN 122023761ACN-122023761-A

Abstract

The method comprises the steps of determining a maximum light intensity difference matrix by determining pulse heights of all pulses in an obtained pulse sequence matrix, wherein the pulse sequence matrix can record time and pulse width corresponding to all pulses of all pixels of an image in a preset time period, and the maximum light intensity difference matrix records the maximum light intensity difference of all the pixels; the method comprises the steps of performing threshold binarization on a processed maximum light intensity difference matrix through an adaptively calculated threshold value to obtain a binarized light intensity difference matrix, performing median filtering on the binarized light intensity difference matrix to obtain a filtered matrix, and calculating the target position and time through taking the intersection of the filtered matrix and the binarized light intensity difference matrix. The method can capture the micro-motion change of the unmanned aerial vehicle target, detect and obtain the position of the unmanned aerial vehicle target in the time period, achieve the effect of detecting the pixel-level target, and greatly improve the detection probability of the unmanned aerial vehicle target with low weak characteristics.

Inventors

CHEN YUPU
HAO WEI
HE SONGLIN
WANG SHUANG

Assignees

航天江南(北京)创新技术研究院有限公司

Dates

Publication Date: 20260512
Application Date: 20251203

Claims (10)

1. A pulse vision-based target detection method, comprising: Determining a maximum light intensity difference matrix by determining the pulse height of each pulse in the obtained pulse sequence matrix, wherein the pulse sequence matrix can record the time and pulse width corresponding to all the pulses of each pixel in a preset time period, and the maximum light intensity difference matrix records the maximum light intensity difference of each pixel; performing threshold binarization on the processed maximum light intensity difference matrix through an adaptively calculated threshold value to obtain a binarized light intensity difference matrix; performing median filtering on the binarized light intensity difference matrix to obtain a filtered matrix; and calculating the target position and time according to the pulse sequence matrix by taking the intersection of the filtered matrix and the binarized light intensity difference matrix.
2. The pulse vision based object detection method of claim 1, wherein after said calculating the object position and time from the pulse sequence matrix by taking the intersection of the filtered matrix and the binarized light intensity difference matrix, comprising: and predicting the motion trail of the target by interpolation according to the target position and the moment.
3. The pulse vision based object detection method according to claim 1, wherein before the determining the maximum light intensity difference matrix by determining the pulse height of each pulse in the acquired pulse sequence matrix, comprising: Acquiring hexadecimal pulse data output by a pulse camera; And converting the hexadecimal pulse data into a decimal pulse sequence matrix.
4. The pulse vision based target detection method according to claim 1, wherein the determining the maximum light intensity difference matrix by determining the pulse height of each pulse in the acquired pulse sequence matrix comprises: calculating the pulse height of each pulse in the pulse sequence matrix; determining the maximum light intensity difference of each pixel according to the pulse height of each pulse; And determining the maximum light intensity difference matrix according to the maximum light intensity difference of each pixel.
5. The pulse vision-based object detection method according to claim 1, wherein before the threshold value through adaptive calculation is used for performing threshold binarization on the processed maximum light intensity difference matrix to obtain a binarized light intensity difference matrix, the method comprises: Quantizing the maximum light intensity difference matrix according to the standard range of the gray value to obtain a quantized light intensity difference matrix; And determining the threshold according to the maximum light intensity difference and the next-largest light intensity difference in the quantized light intensity difference matrix.
6. The pulse vision based target detection method according to claim 5, wherein the performing thresholding on the processed maximum light intensity difference matrix by the adaptively calculated threshold value to obtain a binarized light intensity difference matrix comprises: and determining the light intensity difference which is not smaller than the threshold value in the quantized light intensity difference matrix as a first preset value, and determining the light intensity difference which is smaller than the threshold value in the quantized light intensity difference matrix as a second preset value so as to obtain the binarized light intensity difference matrix.
7. The pulse vision based target detection method of claim 6, wherein the binarized light intensity difference matrix a b is: The position of the pixel in the image is (i, j), A is a quantized light intensity difference matrix, the first preset value is 1, the second preset value is 0, and T is a threshold value.
8. The pulse vision based target detection method of claim 1, wherein the filtered matrix a m is: A m ＝med{A b (i-m,j-n),(m,n∈w)} wherein A b is a binarized light intensity difference matrix, the position of a pixel in the image is (i, j), the offset of the pixel in the image is (m, n), and w is the window size of median filtering.
9. The pulse vision based target detection method of claim 7, wherein said calculating the target position and time from the pulse sequence matrix by taking the intersection of the filtered matrix and the binarized light intensity difference matrix comprises: taking the intersection of the filtered matrix and the binarized light intensity difference matrix to obtain an intersection matrix; determining non-zero element points in the intersection matrix; according to the pixel positions corresponding to the non-zero element points in the intersection matrix, the pixel positions are target positions; converting the pulse sequence matrix into a matrix recorded in the form of light intensity; and determining the corresponding moment when the light intensity in the matrix recorded in the light intensity form is maximum at the same pixel position, namely the moment corresponding to the target position.
10. A pulse vision-based object detection system, comprising: The computing module is used for determining a maximum light intensity difference matrix by determining the pulse height of each pulse in the obtained pulse sequence matrix, wherein the pulse sequence matrix can record the time and the pulse width corresponding to all the pulses of each pixel in a preset time period, and the maximum light intensity difference matrix records the maximum light intensity difference of each pixel; the binarization module is used for carrying out threshold binarization on the processed maximum light intensity difference matrix through the adaptively calculated threshold value to obtain a binarized light intensity difference matrix; The filtering module is used for carrying out median filtering on the binarized light intensity difference matrix to obtain a filtered matrix; And the detection module is used for calculating the target position and time according to the pulse sequence matrix by taking the intersection of the filtered matrix and the binarized light intensity difference matrix.

Description

Target detection method and system based on pulse vision Technical Field The invention relates to the technical field of target detection, in particular to a target detection method and system based on pulse vision. Background The current detection means for unmanned aerial vehicles mainly comprise radar detection, radio frequency monitoring, visual detection and other technologies. The radar detection distance is long, the accuracy is high, unmanned aerial vehicles in a radio silence flight state can be detected, unmanned aerial vehicles in a hovering state cannot be detected, and the unmanned aerial vehicles are easily interfered by low-altitude ground clutter and have detection blind areas. The radio frequency monitoring passive receiving signal has small influence by the environment and good concealment, but can not provide the position information of the unmanned aerial vehicle. The vision detection system captures drone vision data, such as images or videos, using camera sensors and then detects the drone therein using a computer vision-based object detection algorithm. Visual detection is easy to identify unmanned aerial vehicle and other objects, has high-precision positioning capability, is visual in detection effect, and is commonly used for evidence obtaining and checking unmanned aerial vehicle in practical application, thus being an indispensable unmanned aerial vehicle detection technology. The detection algorithm based on vision is researched by mostly adopting a deep learning model, such as the detection method of the unmanned aerial vehicle based on visible light and infrared images, provided by CN202410220340.9, and the detection precision of the unmanned aerial vehicle under different modes is improved through multi-mode unmanned aerial vehicle detection and training. However, in the actual combat environment, due to numerous factors such as illumination, target movement speed and shielding, the imaging effect of a small target under a complex background is poor, and the visible light and infrared visual detection modes are difficult to capture unmanned aerial vehicle targets with strong maneuverability under long distances. Disclosure of Invention The invention aims to provide a target detection method based on pulse vision, which can overcome the defects existing in the conventional vision detection algorithm. In order to achieve the above purpose, the invention adopts the following technical scheme: in one aspect, the present disclosure provides a method for detecting a target based on pulse vision, including: Determining a maximum light intensity difference matrix by determining the pulse height of each pulse in the obtained pulse sequence matrix, wherein the pulse sequence matrix can record the time and pulse width corresponding to all the pulses of each pixel in a preset time period, and the maximum light intensity difference matrix records the maximum light intensity difference of each pixel; performing threshold binarization on the processed maximum light intensity difference matrix through an adaptively calculated threshold value to obtain a binarized light intensity difference matrix; performing median filtering on the binarized light intensity difference matrix to obtain a filtered matrix; and calculating the target position and time according to the pulse sequence matrix by taking the intersection of the filtered matrix and the binarized light intensity difference matrix. In some embodiments of the present disclosure, after the calculating the target position and time according to the pulse sequence matrix by taking the intersection of the filtered matrix and the binarized light intensity difference matrix, the method includes: and predicting the motion trail of the target by interpolation according to the target position and the moment. In some embodiments of the present disclosure, before determining the maximum light intensity difference matrix by determining the pulse height of each pulse in the acquired pulse sequence matrix, the method includes: Acquiring hexadecimal pulse data output by a pulse camera; And converting the hexadecimal pulse data into a decimal pulse sequence matrix. In some embodiments of the present disclosure, the determining the maximum light intensity difference matrix by determining a pulse height of each pulse in the acquired pulse sequence matrix includes: calculating the pulse height of each pulse in the pulse sequence matrix; determining the maximum light intensity difference of each pixel according to the pulse height of each pulse; And determining the maximum light intensity difference matrix according to the maximum light intensity difference of each pixel. In some embodiments of the present disclosure, before the adaptively calculating the threshold, performing threshold binarization on the processed maximum light intensity difference matrix to obtain a binarized light intensity difference matrix, the method includes: Quantizing the maximum l