An intelligent detection method for slow and small targets at sea based on environment perception

Abstract

The application provides an offshore slow small target intelligent detection method based on environment perception. Radar single frame image processing is performed in one antenna scanning period as a frame, environment perception is performed in real time, the current environment state is determined first, and then region division is performed according to the current environment state, then different region detection thresholds are set, and finally target detection is performed in a sliding window mode. The application can better overcome the problems that the traditional constant false alarm detection method is redundant in calculation and is easily affected by the environment, can reduce the calculation amount while maximally reducing the influence of the elevated thresholds of cloud rain clutter, sea clutter and the like on the detection of slow small targets in the non-clutter region, and has a good offshore slow small target intelligent detection effect.

Description

Technical Field

[0001] This invention belongs to the field of radar technology, specifically relating to an intelligent detection method for small, slow-moving targets at sea based on environmental perception. Background Technology

[0002] The detection of small, slow-moving targets at sea is of great significance for both civilian and military applications. Due to the complex, windy, and rainy environment at sea, the detection of such targets is easily affected by cloud and rain clutter and sea clutter. Traditional sea target detection methods based on the constant false alarm rate (CFAR) theory do not make specific adaptive adjustments for the real-time environment, resulting in significant computational redundancy through cell-by-cell calculations. This is especially problematic for maritime radar, where once the radar is installed, the scene is essentially fixed. Under good weather conditions, the background echo across the entire frame does not fluctuate significantly, making the CFAR detection method computationally overly redundant in such scenarios. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention provides an intelligent detection method for small, slow-moving targets at sea based on environmental perception. A single-frame radar image is processed using one antenna scan cycle as a frame, and environmental perception is performed in real time. First, the current environmental state is determined, and regions are divided accordingly. Then, different detection thresholds are set for different regions. Finally, a sliding window approach is used for target detection. This invention effectively overcomes the computational redundancy and susceptibility to environmental influences of traditional constant false alarm rate (CFAR) detection methods. It can reduce computational load while minimizing the impact of cloud and rain clutter, sea clutter, and other factors that raise the threshold on the detection of small, slow-moving targets in clutter-free areas, resulting in superior intelligent detection performance for small, slow-moving targets at sea.

[0004] A method for intelligent detection of small, slow-moving targets at sea based on environmental perception, characterized by the following steps: Step 1: Power on the radar, initialize it, and start transmitting to acquire frame data; Step 2: Calculate environmental parameters through whole-frame sampling integration and multi-frame smoothing operations. The details are as follows: (1) (2) in, This is the mean of the sampling integrals for a single frame. The azimuth dimension of the entire radar image is sampled at 1° intervals. The range dimension of the entire radar image is sampled at 1km intervals. For the first image sampling data i s OK j s The amplitude of the resolution unit of the column, The number of frames, with a value range of K≥10; Step 3: Calculate the reference values ​​for environmental parameter changes using the following formulas. Measured environmental parameter changes and rate of change : (3) (4) (5) in, The environmental parameter baseline value is one. The second environmental parameter baseline value; if and If cloud clutter or sea clutter is detected, proceed to step 4; otherwise, if only one normal region is detected, set the number of regions Z=1 and proceed to step 5. The environmental parameter baseline value one and environmental parameter baseline value 2 The data was obtained through radar pre-acquisition mode, i.e., the radar was operated during times of clear weather and good sea conditions to collect frame data, and the environmental parameter baseline value was calculated through whole-frame sampling integration and multi-frame smoothing. The radar was operated during times of poor weather and sea conditions to collect frame data, and the environmental parameter baseline value was calculated through whole-frame sampling integration and multi-frame smoothing. ; Step 4: Use the directional gradient method to detect edge points in the image, form edge contours, and label areas with contour size greater than or equal to 1km that are not island or coastline contours. Step 5: Calculate the detection threshold for each region using the following formula: (6) in, VT a Indicates the first a Detection thresholds for each region a =1,…,Z, where Z is the number of regions. N a Indicates the first a The number of resolvable units contained in each region. Zone ( a ) indicates the first a Images of each region, A ij For the first a In the image of the region, the first i OK j The amplitude of the column's resolving unit; Step 6: Perform target detection based on the detection threshold set in Step 5, specifically as follows: For any distance dimension in the image, a sliding window detection is performed using J orientation units as a window. If the amplitude of Q resolvable units within the window is greater than or equal to the detection threshold value for that region, then the detection is successful. VT If the target echo is detected within this window, then the detection continues until the entire frame is detected, thus completing the target detection. Here, the value of J ranges from 5 ≤ J ≤ 10, and Q = .

[0005] Specifically, the island or coastline outline mentioned in step 4 can be obtained by marking the acquired frame image of the environmental parameter baseline value.

[0006] The beneficial effects of this invention are as follows: By sensing environmental information in real time and classifying different environmental conditions, the impact of sea surface environmental changes such as clouds, rain, and waves on target detection results can be minimized; by using whole-frame sampling integration and multi-frame smoothing operations to calculate environmental parameters and using their rate of change for judgment, clutter regions and normal regions can be accurately divided, laying a good foundation for adaptive threshold detection and improving target detection accuracy; by using different adaptive threshold detection methods for different regions, intelligent target detection can be achieved with better detection results; this invention is computationally simple, reducing computational load while minimizing the impact of cloud, rain, clutter, and sea clutter raising the threshold on the detection of slow and small targets in clutter-free regions, resulting in good intelligent detection performance for slow and small targets at sea. Attached Figure Description

[0007] Figure 1 This is a flowchart of an intelligent detection method for small, slow-moving targets at sea based on environmental perception, according to the present invention. Detailed Implementation

[0008] The present invention will be further described below with reference to the accompanying drawings and embodiments. The present invention includes, but is not limited to, the following embodiments.

[0009] like Figure 1 As shown, this invention provides an intelligent detection method for small, slow-moving targets at sea based on environmental perception. The specific implementation process is as follows: Step 1: Power on the radar, initialize it, and start transmitting to acquire frame data.

[0010] Step 2: Calculate environmental parameters through whole-frame sampling integration and multi-frame smoothing operations. The details are as follows: (7) (8) in, This is the mean of the sampling integrals for a single frame. The azimuth dimension of the entire radar image is sampled at 1° intervals. The range dimension of the entire radar image is sampled at 1km intervals. For the first image sampling data i s OK j s The amplitude of the resolution unit of the column, The frame number is K, and its value range is K≥10.

[0011] Step 3: Calculate the reference values ​​for environmental parameter changes using the following formulas. Measured environmental parameter changes and rate of change : (9) (10) (11) in, As the environmental parameter baseline value one, The second environmental parameter benchmark value is used to more accurately determine the current environmental state. This invention uses the environmental benchmark value for environmental determination, selecting two environmental states as benchmarks: State 1 is a time of clear weather and good sea conditions, and State 2 is a time of poor weather and poor sea conditions. Frame data is acquired by radar pre-collection during these two times, and then the environmental parameter benchmark value is calculated through whole-frame sampling integration and multi-frame smoothing operations (formulas (7) and (8)). Specifically, the radar is operated during a time of clear weather and good sea conditions to collect frame data, and the environmental parameter benchmark value is calculated through whole-frame sampling integration and multi-frame smoothing operations. The radar was operated during times of poor weather and sea conditions to collect frame data. Environmental parameter baseline values ​​were then calculated through whole-frame sampling integration and multi-frame smoothing operations. .

[0012] After calculating the environmental parameters and environmental change rate as described above, an environmental assessment is performed to delineate different regions, i.e., if... and If cloud / rain clutter or sea clutter is detected, proceed to step 4; otherwise, if only one normal region is detected, set the number of regions Z=1 and proceed to step 5.

[0013] Step 4: Use the directional gradient method to detect edge points in the image and form edge contours. Label regions with contour sizes greater than or equal to 1 km that are not island or coastline contours. Assume there are Z regions in total (including cloud / rain clutter and normal areas). Island or coastline contours can be obtained beforehand by marking the acquired frame images of the environmental parameter baseline value.

[0014] Step 5: Calculate the detection threshold for each region using the following formula: (12) in, VT a Indicates the first a Detection thresholds for each region a =1,…,Z, where Z is the number of regions. N a Indicates the first a The number of resolvable units contained in each region. Zone ( a ) indicates the first a Images of each region, A ij For the first a In the image of the region, the first i OK j The amplitude of the resolution unit of the column.

[0015] Step 6: Perform target detection based on the detection threshold set in Step 5, specifically as follows: For any distance dimension in the image, a sliding window detection is performed using J orientation units as a window. If the amplitude of Q resolvable units within the window is greater than or equal to the detection threshold value for that region, then the detection is successful. VT If the target echo is detected within this window, then the detection continues until the entire frame is detected, thus completing the target detection. Here, the value of J ranges from 5 ≤ J ≤ 10, and Q = .

[0016] This invention abandons the traditional constant false alarm rate (CFAR) target detection method. Instead, it processes radar images in a single frame, with one antenna scan cycle as one frame, and performs real-time environmental perception. If cloud, rain, or sea clutter is present, the clutter location is determined through differential operations. The entire radar image frame is divided into clutter and normal regions, and different adaptive thresholds are applied to different regions to minimize the impact of cloud, rain, or sea clutter on the detection of small, slow targets in clutter-free areas. Compared to the CFAR detection method, this invention has a lower computational load, and with subsequent inter-frame detection processing, the overall detection performance is improved by 3-5 dB compared to traditional CFAR detection methods.

Claims

1. A method for intelligent detection of small, slow-moving targets at sea based on environmental perception, characterized in that... The steps are as follows: Step 1: Power on the radar, initialize it, and start transmitting to acquire frame data; Step 2: Calculate environmental parameters through whole-frame sampling integration and multi-frame smoothing. The details are as follows: （1） （2） in, This is the mean of the sampling integrals for a single frame. The azimuth dimension of the entire radar image is sampled at 1° intervals. The range dimension of the entire radar image is sampled at 1km intervals. For the first image sampling data i s OK j s The amplitude of the resolution unit of the column, The number of frames, with a value range of K≥10; Step 3: Calculate the reference values ​​for environmental parameter changes using the following formulas. Measured environmental parameter changes and rate of change : （3） （4） （5） in, As the environmental parameter baseline value one, The second environmental parameter baseline value; if and If cloud clutter or sea clutter is detected, proceed to step 4; otherwise, if only one normal region is detected, set the number of regions Z=1 and proceed to step 5. The environmental parameter baseline value one and environmental parameter baseline value 2 The data was obtained through radar pre-acquisition mode, i.e., the radar was operated during times of clear weather and good sea conditions to collect frame data, and the environmental parameter baseline value was calculated through whole-frame sampling integration and multi-frame smoothing. The radar was operated during times of poor weather and sea conditions to collect frame data. Environmental parameter baseline values ​​were then calculated through whole-frame sampling integration and multi-frame smoothing operations. ; Step 4: Use the directional gradient method to detect edge points in the image, form edge contours, and label areas with contour size greater than or equal to 1km that are not island or coastline contours. Step 5: Calculate the detection threshold for each region using the following formula: （6） in, VT a Indicates the first a Detection thresholds for each region a =1,…,Z, where Z is the number of regions. N a Indicates the first a The number of resolvable units contained in each region. Zone ( a ) indicates the first a Images of each region, A ij For the first a In the image of the region, the first i OK j The amplitude of the column's resolving unit; Step 6: Perform target detection based on the detection threshold set in Step 5, specifically as follows: For any distance dimension in the image, a sliding window detection is performed using J orientation units as a window. If the amplitude of Q resolvable units within the window is greater than or equal to the detection threshold value for that region, then the detection is successful. VT If the target echo is detected within this window, then the detection continues until the entire frame is detected, thus completing the target detection. Here, the value of J ranges from 5 ≤ J ≤ 10, and Q = .

2. The intelligent detection method for small, slow-moving targets at sea based on environmental perception as described in claim 1, characterized in that: The island or coastline outline mentioned in step 4 can be obtained by marking the acquired frame image of the environmental parameter baseline value 1.

Images