Tower footing hyperspectral semantic change detection method considering illumination change

Abstract

The invention relates to a semantic change detection method, in particular to a tower footing hyperspectral semantic change detection method considering illumination change. The method comprises the following steps: S1, determining a target scene according to the position of a tower footing remote sensing platform and an optical path coverage range, and presetting a hyperspectral imager; moving the at least three gray level targets with gradient differences to the target scene; s2, acquiring image data of the target scene and all the gray scale targets by using a hyperspectral imager, and acquiring radiance values of all the gray scale targets according to the image data of all the gray scale targets; performing radiation correction on the target scene by using the radiance values of all the gray scale targets and the known reflectivity values thereof to obtain the radiance value of the target scene; and S3, performing semantic change detection analysis by using the hyperspectral data subjected to radiation correction. According to the invention, radiation correction of the target scene can be realized, and more accurate surface feature reflectivity data can be obtained.

Description

Technical Field

[0001] The invention relates to a semantic change detection method, and in particular to a tower base hyperspectral semantic change detection method considering illumination changes. Background Art

[0002] In the field of remote sensing monitoring, hyperspectral imaging technology is widely used in scenarios such as object identification, environmental monitoring, and resource exploration because it can provide rich spectral information. However, changes in lighting conditions, such as solar altitude, atmospheric conditions, and cloud cover, can have a significant impact on the acquisition of hyperspectral data, resulting in differences in the spectral reflectance of the same object under different lighting conditions. This difference may be misjudged as a change in the object itself, thus affecting the accuracy of change detection.

[0003] Existing hyperspectral change detection methods usually rely on comparing hyperspectral image data acquired at different time points to identify changes in ground objects. However, these methods often do not fully consider the impact of illumination changes on spectral data, resulting in false positives or omissions in change detection results when illumination conditions change significantly. In addition, since the tower-based hyperspectral imaging system is usually fixed at a specific location, its observation conditions are relatively stable, but changes in illumination conditions are still inevitable during long-term monitoring.

[0004] In order to improve the accuracy of change detection, it is urgent to develop a tower base hyperspectral change detection method that can take into account the impact of illumination changes. Summary of the invention

[0005] The purpose of the present invention is to solve the technical problem that the existing detection methods are prone to inaccurate detection results due to spectral differences caused by illumination changes and actual changes in ground objects, and to provide a tower base hyperspectral semantic change detection method that takes illumination changes into consideration.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is:

[0007] A tower base hyperspectral semantic change detection method considering illumination changes is special in that it includes the following steps:

[0008] S1. Determine the target scene according to the position and optical path coverage of the tower base remote sensing platform, and pre-set a hyperspectral imager that can rotate 360°. Adjust the hyperspectral imager to ensure that the imaging clarity meets the set requirements; at the same time, move at least three grayscale targets with gradient differences to the target scene, so that the hyperspectral imager can simultaneously obtain image data of the target scene and the grayscale targets;

[0009] S2. Use a hyperspectral imager to collect image data of the target scene and all gray-scale targets, and obtain the radiance values ​​of all gray-scale targets based on the image data of all gray-scale targets; and use the radiance values ​​of all gray-scale targets and their known reflectance values ​​to perform radiation correction on the target scene to obtain the radiance value of the target scene, that is, hyperspectral data;

[0010] S3. Using radiometrically corrected hyperspectral data for semantic change detection analysis

[0011] S3.1, preprocess the hyperspectral data after radiation correction;

[0012] S3.2, input the preprocessed hyperspectral data into the trained twin neural network, learn and compare the hyperspectral data at two different time points through the twin neural network, and obtain the BCD map and the semantic map at two different time points;

[0013] S3.3. Obtain the SCD map based on the BCD map and the semantic maps at two different time points to complete the tower base hyperspectral semantic change detection.

[0014] Furthermore, in S2, using the radiance values ​​of all grayscale targets and their known reflectivity values, the transformation coefficients A and C are first calculated by the following formula:

[0015] L a =Gain×DN+Bias

[0016] L a =A·ρ a +C

[0017] Where: L a is the radiance value of the grayscale target, Gain is the gain of the hyperspectral imager, DN is the digital quantization value output by each detection element of the hyperspectral imager, Bias is the deviation value of the hyperspectral imager, ρ a is the surface reflectivity of the grayscale target;

[0018] Then, using the calculated transformation coefficients A and C, the radiance value L of the target scene is calculated by the following formula: tar :

[0019] L tar =A·ρ tar +C

[0020] Among them, ρ tar is the surface reflectivity of the target scene. ,

[0021] Furthermore, the size of the hyperspectral data is adjusted to match the input requirements of the twin neural network, and the adjusted hyperspectral data is normalized to complete the preprocessing of the hyperspectral data after radiation correction.

[0022] Furthermore, in S3.2, based on the hyperspectral data at two different time points, the deep BCD features and the deep semantic features at two different time points are obtained through the twin neural network;

[0023] According to the deep semantic features at two different time points, the semantic graphs at two different time points are obtained; at the same time, the deep semantic features at two different time points are fused with the deep BCD features to obtain the BCD graph.

[0024] Furthermore, in S3.3, the SCD graph is obtained through mask operation based on the semantic graphs of the BCD graph at two different time points.

[0025] Furthermore, it also includes S4, analyzing and verifying the SCD graph to evaluate the accuracy of the tower base hyperspectral semantic change detection.

[0026] The beneficial effects of the present invention are:

[0027] The tower base hyperspectral semantic change detection method considering illumination changes proposed in the present invention significantly improves the accuracy of hyperspectral semantic change detection through precise radiation correction and advanced twin neural networks, especially for the inconsistency of ground object radiation at different times and under different weather conditions, and realizes radiation correction of the target scene, thereby obtaining more accurate ground object reflectance data, effectively reducing false alarms and missed alarms, and enhancing the reliability of remote sensing monitoring results. The rich dimensions and coherent imaging time of the collected data enable the present invention to achieve long-term automated unattended observation, meet the data requirements of different time scales, and significantly improve the accuracy of change detection and the continuity of monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic diagram of a radiation correction method of a tower base hyperspectral semantic change detection method considering illumination changes of the present invention;

[0029] Figure 2 It is a flow chart of the tower base hyperspectral semantic change detection method considering illumination changes of the present invention;

[0030] Figure 3 It is a flow chart of radiation correction of the tower base hyperspectral semantic change detection method considering illumination changes of the present invention;

[0031] Figure 4 It is a twin neural network flow chart of the tower base hyperspectral semantic change detection method considering illumination changes in the present invention. DETAILED DESCRIPTION

[0032] In order to make the purpose, advantages and features of the present invention more clear, the following is a further detailed description of a tower base hyperspectral semantic change detection method considering illumination changes proposed by the present invention in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will be more clear according to the following specific implementation methods.

[0033] See also Figure 2 The present embodiment provides a tower base hyperspectral semantic change detection method considering illumination changes, which specifically includes the following steps:

[0034] S1. See Figure 1 , using the positioning capability of the tower-based remote sensing platform, determine the target scene according to the position and optical path coverage of the tower-based remote sensing platform, and adjust the hyperspectral imager to ensure that the imaging clarity meets the set requirements, so as to clearly capture the target scene within a range of 10 kilometers; at the same time, move at least three grayscale targets with gradient differences to the target scene, so that the hyperspectral imager can simultaneously obtain the image data of the target scene and the grayscale target, providing basic data for subsequent analysis and processing.

[0035] S2. Given the reflectivity of all gray-scale targets, use a hyperspectral imager to collect image data of the target scene and all gray-scale targets to obtain the radiance values ​​of all gray-scale targets. Using the radiance values ​​of all gray-scale targets and their known reflectivity values, first calculate the transformation coefficients A and C using the following formula:

[0036] L a =Gain×DN+Bias

[0037] L a =A·ρ a +C

[0038] Where: L a is the radiance value of the grayscale target, Gain is the gain of the hyperspectral imager, DN is the digital quantization value output by each detection element of the hyperspectral imager, Bias is the deviation value of the hyperspectral imager, ρ a is the surface reflectivity of the grayscale target.

[0039] Then, using the calculated transformation coefficients A and C, the radiance value L of the target scene is calculated by the following formula: tar To perform a radiometric correction:

[0040] L tar =A·ρ tar +C

[0041] Among them, ρ tar is the surface reflectivity of the target scene.

[0042] The radiance value of the aforementioned target scene is the hyperspectral data.

[0043] S3, see Figure 3 , using radiometrically corrected hyperspectral data for semantic change detection analysis

[0044] S3.1. Preprocess the hyperspectral data after radiation correction

[0045] The size of the hyperspectral data is adjusted to match the input requirements of the twin neural network, and the adjusted hyperspectral data is normalized to complete the preprocessing of the hyperspectral data after radiation correction.

[0046] S3.2. The preprocessed hyperspectral data is input into the trained twin neural network. The twin neural network can learn and compare the hyperspectral data at two different time points, namely, T1 image and T2 image, and identify the differences between them.

[0047] The twin neural network uses the existing network and trains it in advance to obtain a trained twin neural network. For details, see Figure 4 ,The twin neural network includes a semantic encoder, a variation encoder, a semantic decoder, and a variation decoder.

[0048] According to the T1 image and the T2 image, the T1 deep semantic features and the T2 deep semantic features are obtained through the semantic encoder in the twin neural network, and the deep BCD features are obtained through the variation encoder in the twin neural network.

[0049] S3.3. According to the deep semantic features at two different time points, the semantic map at time T1 and the semantic map at time T2 are obtained through the semantic decoder in the twin neural network. At the same time, the semantic map at time T1 and the semantic map at time T2 are fused with the deep BCD features through the twin neural network, and the fused features are decoded through the change decoder in the twin neural network to obtain the BCD map (binary change map); then, according to the BCD map and the deep semantic features at two different time points, the SCD map (semantic change map) is obtained through masking operation to complete the hyperspectral semantic change detection of the tower base.

[0050] S4. The SCD graph is analyzed and verified to evaluate the accuracy of the tower base hyperspectral semantic change detection.

Claims

1. A tower base hyperspectral semantic change detection method considering illumination changes, characterized in that: The following steps are involved: S1. Determine the target scene according to the position and optical path coverage of the tower-based remote sensing platform, and adjust the hyperspectral imager to ensure that the imaging clarity meets the set requirements; at the same time, move at least three grayscale targets with gradient differences to the target scene so that the hyperspectral imager can simultaneously obtain image data of the target scene and the grayscale targets; S2. Use a hyperspectral imager to collect image data of the target scene and all gray-scale targets, and obtain the radiance values ​​of all gray-scale targets based on the image data of all gray-scale targets; and use the radiance values ​​of all gray-scale targets and their known reflectance values ​​to perform radiation correction on the target scene to obtain the radiance value of the target scene, that is, hyperspectral data; S3. Using radiometrically corrected hyperspectral data for semantic change detection analysis S3.1, preprocess the hyperspectral data after radiation correction; S3.2, input the preprocessed hyperspectral data into the trained twin neural network, learn and compare the hyperspectral data at two different time points through the twin neural network, and obtain the BCD map and the semantic map at two different time points; S3.

3. Obtain the SCD map based on the BCD map and the semantic maps at two different time points to complete the tower base hyperspectral semantic change detection.

2. The tower foundation hyperspectral semantic change detection method considering illumination changes according to claim 1 is characterized in that: In S2, using the radiance values ​​of all gray-scale targets and their known reflectance values, the transformation coefficients A and C are first calculated using the following formula: L a =Gain×DN+Bias L a =A·ρ a +C Where: L a is the radiance value of the grayscale target, Gain is the gain of the hyperspectral imager, DN is the digital quantization value output by each detection element of the hyperspectral imager, Bias is the bias value of the hyperspectral imager, ρ a is the surface reflectivity of the grayscale target; Then, using the calculated transformation coefficients A and C, the radiance value L of the target scene is calculated by the following formula: tar : L tar =A·ρ tar +C Among them, ρ tar is the surface reflectivity of the target scene.

3. The tower foundation hyperspectral semantic change detection method considering illumination changes according to claim 1 or 2, characterized in that: S3.1 specifically: The size of the hyperspectral data is adjusted to match the input requirements of the twin neural network, and the adjusted hyperspectral data is normalized to complete the preprocessing of the hyperspectral data after radiation correction.

4. The tower foundation hyperspectral semantic change detection method considering illumination changes according to claim 3 is characterized in that: In S3.2, based on the hyperspectral data at two different time points, the deep BCD features and the deep semantic features at two different time points are obtained through the twin neural network; According to the deep semantic features at two different time points, the semantic graphs at two different time points are obtained; at the same time, the deep semantic features at two different time points are fused with the deep BCD features to obtain the BCD graph.

5. The tower foundation hyperspectral semantic change detection method considering illumination changes according to claim 4 is characterized in that: In S3.3, the SCD graph is obtained through mask operation based on the BCD graph and the semantic graphs at two different time points.

6. The tower foundation hyperspectral semantic change detection method considering illumination changes according to claim 1 is characterized in that: Also included is S4, which analyzes and validates the SCD graph to evaluate the accuracy of the hyperspectral semantic change detection in the tower.

Images