阅读说明：本技术 一种软边缘支持向量机仿生偏振光航向解算系统及方法 (Bionic polarized light course resolving system and method for soft edge support vector machine ) 是由 白宏阳 梁华驹 郭宏伟 郑浦 李政茂 周育新 胡珂 于 2020-04-23 设计创作，主要内容包括：本发明公开了一种软边缘支持向量机仿生偏振光航向解算系统及方法,系统包括水平仪、像素级偏振相机、数据处理计算机、锂电池和安装架,通过水平仪保证安装架上的像素级偏振相机正对天顶点；通过像素级偏振相机实时采集大气偏振模式得到偏振度和偏振角图像；通过数据处理计算机进行航向解算；该方法步骤为：对偏振相机采集的偏振图片进行阈值分割得到两区域图像；对两区域图像进行连通域检测得到三区域图像；对三区域图像的偏振角取反得到可分类的偏振图像；对可分类的偏振角图像利用软边缘支持向量机进行航向解算得到航向角。本发明将航向解算问题转化为二分类问题并利用软边缘支持向量机求解,抗干扰性强,可在多云和阴天条件下进行航向解算。(The invention discloses a bionic polarized light course resolving system and a bionic polarized light course resolving method for a soft edge support vector machine, wherein the system comprises a level meter, a pixel level polarization camera, a data processing computer, a lithium battery and an installation rack, and the pixel level polarization camera on the installation rack is ensured to be over against a zenith point through the level meter; acquiring an atmospheric polarization mode in real time through a pixel-level polarization camera to obtain polarization degree and polarization angle images; calculating course through a data processing computer; the method comprises the following steps: carrying out threshold segmentation on a polarization picture acquired by a polarization camera to obtain two area images; carrying out connected domain detection on the two-region images to obtain a three-region image; inverting the polarization angle of the three-region image to obtain a classifiable polarization image; and carrying out course calculation on the classifiable polarization angle image by using a soft edge support vector machine to obtain a course angle. The invention converts the course resolving problem into a two-classification problem and solves the problem by using a soft edge support vector machine, has strong anti-interference performance and can carry out course resolving under the conditions of cloudy days and cloudy days.)

1. A bionic polarized light course resolving system of a soft edge support vector machine is characterized by comprising a level meter, a pixel level polarization camera, a data processing computer and a lithium battery;

the level meter is used for ensuring that the pixel level polarization camera is over against the zenith point, and the pixel level polarization camera acquires an atmospheric polarization mode in real time to obtain images of polarization degree and polarization angle and transmits the images to the data processing computer through the network port; the data processing computer is used for processing the polarization angle image acquired by the pixel-level polarization camera and carrying out course calculation; the lithium battery supplies power to the level gauge, the pixel level polarization camera and the data processing computer.

2. The bionic polarized light course calculation system of the soft edge support vector machine according to claim 1, wherein the pixel-level polarization camera is provided with four polarizing films in different directions on each pixel array element of the camera, wherein the polarizing films are respectively a 0-degree polarizing film, a 45-degree polarizing film, a 90-degree polarizing film and a 135-degree polarizing film; the real-time collection of the atmospheric polarization mode is realized through the polaroids in four different directions.

3. The bionic polarized light heading solution system of the soft edge support vector machine according to claim 1, further comprising a mounting rack for fixing the level gauge, the lithium battery and the polarization camera.

4. A course solving method of a bionic polarized light course solving system based on the soft edge support vector machine of claim 1 is characterized by comprising the following steps:

step 1, carrying out threshold segmentation on a polarization angle image acquired by a polarization camera to obtain two area images;

step 2, carrying out connected domain detection on the two-region images to obtain a three-region image;

step 3, reversing the polarization angle of the three-region image to obtain a classifiable polarization angle image;

step 4, classifying the classifiable polarization angle images in the step 3 through a soft edge support vector machine, and resolving and converting the course angle into a classification problem;

step 5, solving the classification problem in the step 4 through a soft edge support vector machine;

step 6, training the soft edge support vector machine in the step 5 to obtain a weight vector, and calculating to obtain a solar course angle;

and 7, performing sun motion compensation on the sun azimuth angle obtained in the step 6 to finally obtain a course angle.

5. The bionic polarized light course calculation method of the soft edge support vector machine according to claim 4 is characterized in that the step 1 specifically comprises the following steps: dividing the detected polarization angle image into two regions by threshold segmentation; the polarization angle is greater than-T and less than T as the region r₁The remainder being r₀A region, where T is a threshold; corroding the image to divide the part with the polarization degree of 0 into regions r₁Will be the region r₁Expands outwardly.

6. The soft edge support vector machine bionic polarized light course calculation method as claimed in claim 5, wherein step 2 is implemented by detecting a connected region and using r in step 1₁、r₀The two areas are distinguished to obtain three marked areas r₁，r₂，r₃The polarization image of (a);

step 3 by marking r as described in step 2₂The polarization angles of the regions are inverted, denoted r₃The polarization angle of the region remains unchanged, mark r₂And (4) when the regional course is calculated, the images with the classifiable polarization angles are obtained, wherein the images with the polarization angles larger than 0 are classified into one type, and the images with the polarization angles smaller than 0 are classified into another type.

7. The bionic polarized light heading solution method of the soft edge support vector machine according to claim 4 or 6, wherein the classification problem in the step 4 is a linear classification problem represented by

w^Tx+b＝0

Where w is the weight vector, b is the offset, and x is the hyperplane w^TArgument of 0 x + b。

8. The bionic polarized light course calculation method of the soft edge support vector machine according to claim 7, wherein the soft edge support vector machine in the step 5 is expressed as

s.t.d_k(w^Tx_k+b)≥1-ξ_k,k＝1,2,…,n.

ξ_k≥0,k＝1,2,…,n.

Wherein, ξ_kIs a relaxation variable, C is a penalty parameter, x_kDenotes the k-th training sample, d_kDenotes x_kBelong to the category, s.t. means "make".

9. The bionic polarized light course calculation method of the soft edge support vector machine according to claim 8, wherein the step 6 is to train the soft edge support vector machine in the step 5 to obtain a weight vector w, and obtain a solar course angle:

where θ is the solar azimuth angle, w₁And w₂Representing the first and second components of the weight vector w, respectively.

Technical Field

The invention belongs to the technical field of bionic polarized light navigation, and particularly relates to a bionic polarized light course resolving system and method of a soft edge support vector machine.

Background

Polarization is an inherent property of light, and reflects the change rule of the vibration direction of light along with time. When entering the atmosphere, the unbiased sunlight is subjected to the scattering, absorption and radiation effects of various particles in the atmosphere, so that the transmission direction, the light intensity and the polarization state of the unbiased sunlight are changed, and an atmospheric polarization mode is generated in the sky. Biological studies have found that many insects with compound eye structures can detect atmospheric polarization patterns by using compound eyes, and thus realize navigation and positioning functions. The stable polarization distribution of the atmospheric polarization mode can be well used as an information source in the navigation field. The practical application of bionic polarized light navigation can be effectively promoted by researching the polarized light navigation mechanism of the insect.

The bionic polarized light navigation is a method for detecting an atmospheric polarization mode to perform navigation and positioning by simulating a special compound eye structure of organisms such as sand ants, bees and the like. The bionic polarized light navigation is difficult to be interfered by human factors, has no accumulated error, is not interfered by electromagnetism, belongs to passive navigation, has good concealment and has the advantages which are not possessed by the traditional navigation method. However, the atmospheric polarization mode used for navigation in the biomimetic polarized light navigation is interfered and influenced by weather changes, so that it is important to design a navigation method capable of coping with various complex weather conditions.

However, the existing polarized light navigation method mainly aims at clear weather, and better navigation accuracy can be obtained under ideal clear weather. However, when the weather condition becomes severe, especially when the interference of cloud layers occurs, the navigation accuracy is seriously reduced, and even navigation cannot be performed. Patent CN201910597332.5 discloses a dynamic orientation method and system based on a strapdown polarized light compass, but the method does not consider the influence of cloud and other severe weather. Patent CN201610076299.8 discloses a two-channel bionic polarized light navigator based on a polarization splitting prism and a method thereof, but the system needs two channels and does not consider the influence of bad weather such as cloud.

Disclosure of Invention

The invention aims to provide a soft edge support vector machine bionic polarized light course resolving system and method to meet the practical application requirements of bionic polarized light navigation in sunny days and other severe weathers.

The technical solution for realizing the purpose of the invention is as follows: a bionic polarized light course resolving system of a soft edge support vector machine comprises a level meter, a pixel level polarization camera, a data processing computer and a lithium battery;

the level meter is used for ensuring that the pixel level polarization camera is over against the zenith point, and the pixel level polarization camera acquires an atmospheric polarization mode in real time to obtain images of polarization degree and polarization angle and transmits the images to the data processing computer through the network port; the data processing computer is used for processing the polarization angle image acquired by the pixel-level polarization camera and carrying out course calculation; the lithium battery supplies power to the level gauge, the pixel level polarization camera and the data processing computer.

A bionic polarized light course resolving method of a soft edge support vector machine comprises the following steps:

step 1, carrying out threshold segmentation on a polarization angle image acquired by a pixel-level polarization camera to obtain two area images;

step 2, carrying out connected domain detection on the two-region images to obtain a three-region image;

step 3, reversing the polarization angle of the three-region image to obtain a classifiable polarization angle image;

step 4, classifying the classifiable polarization angle images in the step 3 through a soft edge support vector machine; resolving and converting the course angle into a classification problem;

step 5, solving the classification problem in the step 4 through a soft edge support vector machine;

step 6, training the soft edge support vector machine in the step 5 to obtain a weight vector, and calculating to obtain a solar course angle;

and 7, performing sun motion compensation on the sun azimuth angle in the step 6 to finally obtain a course angle.

Compared with the prior art, the invention has the following remarkable advantages: (1) the four polarizing films in different directions are directly arranged on the camera pixel array element, so that the atmospheric polarization mode can be acquired in real time, and the device is small in size and light in weight; (2) the course angle resolving problem is converted into a two-classification problem, interference of clouds and noise is considered, course resolving is carried out by using a soft edge support vector machine, and therefore not only can bionic polarized light navigation under the condition of clear weather be realized, but also polarized light navigation under severe weather such as cloudy and cloudy days can be realized.

Drawings

FIG. 1 is a schematic structural diagram of a bionic polarized light course resolving system of a soft edge support vector machine of the invention.

FIG. 2 is a structural diagram of a polarization camera of the bionic polarized light course resolving system of the soft edge support vector machine of the invention.

FIG. 3 is a flow chart of a bionic polarized light course resolving method of the soft edge support vector machine of the invention.

Detailed Description

With reference to fig. 1, the bionic polarized light course resolving system of the soft edge support vector machine of the invention comprises a level meter, a pixel level polarization camera, a data processing computer, a lithium battery and a mounting rack;

the gradienter ensures that the pixel-level polarization camera arranged on the mounting rack is right opposite to the zenith point, and collects the atmospheric polarization mode in real time; the pixel-level polarization camera acquires an atmospheric polarization mode in real time to obtain a polarization Degree (DOP) image and a polarization Angle (AOP) image, and transmits the images to the data processing computer through a network port; the data processing computer processes the polarization image acquired by the pixel-level polarization camera and carries out course calculation; the lithium battery supplies power to the level meter, the pixel level polarization camera and the data processing computer; the mounting bracket is used for fixing the level gauge, the lithium battery and the polarization camera.

Furthermore, the pixel-level polarization camera is directly provided with four polarizing plates in different directions on a camera pixel array element, wherein the polarizing plates are respectively a 0-degree polarizing plate, a 45-degree polarizing plate, a 90-degree polarizing plate and a 135-degree polarizing plate; and the real-time collection of the atmospheric polarization mode is realized through the polarizing films in the four different directions.

The invention also provides a bionic polarized light course resolving method of the soft edge support vector machine based on the system, which is used for converting the course resolving problem into a two-classification problem and solving the problem by using the soft edge support vector machine aiming at the interference of severe weather and comprises the following steps:

step 1, dividing the detected polarization angle image into two regions by threshold segmentation. The polarization angle is greater than-T and less than T as the region r₁The remainder being r₀A region, where T is a threshold; corroding the image to eliminate the interference of noise; divided into regions r by portions having a degree of polarization of 0₁Eliminating the interference of local cloud; passing region r₁And the expansion is outward, and the interference of cloud boundaries is eliminated.

Step 2, through the detection of the communicated region, the two regions in the step 1 are distinguished to obtain three regions r with marks₁，r₂，r₃The polarization image of (a).

Step 3, marking the mark r in the step 2₂The polarization angles of the regions are inverted, denoted r₃The polarization angle of the region remains unchanged, mark r₂And (4) when the regional heading angle is calculated, the images with the classifiable polarization angles are obtained, wherein the polarization angle is larger than 0, and the polarization angle is smaller than 0.

And 4, classifying the classifiable image in the step 3 through a soft edge support vector machine. And resolving and converting the course angle into a classification problem. The classification problem can be expressed as

w^Tx+b＝0

Where w is the weight vector, b is the offset, and x is the hyperplane w^Tx + b is an argument of 0.

Step 5, solving the classification problem in the step 4 through a soft edge support vector machine, wherein the soft edge support vector machine further considers the interference of noise and cloud, and can effectively cope with severe weather; the soft edge support vector machine may be represented as:

s.t.d_k(w^Tx_k+b)≥1-ξ_k，k＝1，2,…,n.

ξ_k≥0，k＝1,2,…，n.

wherein, ξ_kIs a relaxation variable, C is a penalty parameter, x_kDenotes the k-th training sample, d_kDenotes x_kBelong to the category, s.t. means "make".

Step 6, training the separable offset angle image in the step 3 through the soft edge support vector machine in the step 5 to obtain a weight vector w, and finally obtaining the solar course angle

Where θ is the sun heading angle, w₁And w₂Representing the first and second components of the weight vector w, respectively.

And 7, performing sun motion compensation on the sun azimuth angle in the step 6 to finally obtain a course angle.

The invention is described in further detail below with reference to the figures and specific embodiments.