扩展角锥波前探测器动态范围的解析方案【附代码】

✨ 长期致力于自适应光学、液晶波前校正器、大气湍流、角锥波前探测器研究工作，擅长数据搜集与处理、建模仿真、程序编写、仿真设计。
✅ 专业定制毕设、代码
✅ 如需沟通交流，点击《获取方式》

---

（1）快反镜调制型角锥探测器动态范围扩展机理：

针对传统角锥探测器动态范围小于1λ的限制，提出在锥尖前引入快反镜使光斑沿圆形轨迹运动，运动半径可调。入射波前倾斜导致光斑圆轨迹中心偏离棱镜锥尖，光斑在各锥面上的曝光时间与偏离量成正比，从而调制四个光瞳像的亮度。调制频率设定为200Hz，轨迹半径设为3倍艾里斑直径。理论分析表明动态范围可从1λ扩展至12λ（λ=780nm）。推导出响应信号与波前斜率的线性关系仍成立，仅需重新标定响应矩阵。在Zemax中仿真验证，当输入倾斜像差8λ时，传统探测器只有一个光瞳像可见，而调制型四个光瞳像亮度差异在可探测范围内（最暗光瞳亮度不低于最亮光瞳的15%）。

（2）光瞳像精确提取与像差探测精度提升：

针对调制过程引入的光瞳像间亮带问题，调节光学系统使光瞳像中心距为光瞳像半径的3.8倍，此时亮带对信号的影响可忽略。采用双角锥棱镜替代单棱镜，消除色散导致的光瞳像边缘模糊，色差从0.15mm降至0.02mm。光瞳像定位采用两级方法：粗定位基于边缘亮度梯度极大值，精定位使用相关算法，生成包含10个同心圆的靶图投射到液晶校正器，计算光瞳像中心坐标在粗定位邻域内振荡时的互相关度，定位精度达到亚像素（0.15像素）。开环探测误差从0.07λ RMS降至0.03λ RMS，闭环校正残差减小15%。

（3）大像差输入下响应矩阵重构与开环自适应光学验证：

传统响应矩阵计算基于小像差近似，当输入像差超过2λ时误差急剧增大。重构方法将调制引入的衰减项建模为调制多项式，通过坐标归一化抑制三角函数振荡。输入像差10λ时，计算35项Zernike模式的响应矩阵，条件数从原本的1780降至32。采用该响应矩阵进行波前重构，探测误差RMS小于λ/20。搭建验证光路，使用Zygo干涉仪产生已知像差（RMS 1.2λ），角锥探测器与干涉仪测量结果差值RMS 0.038λ。最终在开环自适应光学系统中测试，校正后斯特列尔比从0.12提升至0.79，接近闭环系统性能（0.85）。

import numpy as np
from scipy.linalg import pinv
from scipy.signal import correlate2d
import cv2

class ModulatedPyramidWFS:
    def __init__(self, pupil_size=64, n_pupils=4, modulation_radius=3.0):
        self.size = pupil_size
        self.n = n_pupils
        self.radius = modulation_radius
    def simulate_modulation(self, tilt_x, tilt_y):
        # shift of circular trajectory center
        dx = tilt_x * 0.5  # conversion factor
        dy = tilt_y * 0.5
        # compute exposure time weights per facet
        weights = np.array([dx+dy, -dx+dy, -dx-dy, dx-dy]) + 1.0
        weights = np.clip(weights / weights.sum(), 0.05, 0.95)
        return weights
    def compute_intensity(self, phase_map, weights):
        # simple geometric optics simulation
        intensity = np.zeros((self.size, self.size, self.n))
        for i in range(self.n):
            intensity[:,:,i] = weights[i] * (1 + 0.5 * np.gradient(phase_map)[0]**2)
        return intensity

def pupil_coregistration(pupil_images, reference_pattern):
    # subpixel alignment using cross-correlation
    best_shifts = []
    for img in pupil_images:
        corr = correlate2d(img, reference_pattern, mode='same')
        peak = np.unravel_index(np.argmax(corr), corr.shape)
        # parabolic fit for subpixel
        y, x = peak
        if 0 < x < corr.shape[1]-1 and 0 < y < corr.shape[0]-1:
            dx = (corr[y, x+1] - corr[y, x-1]) / (2*corr[y, x] - corr[y, x-1] - corr[y, x+1])
            dy = (corr[y+1, x] - corr[y-1, x]) / (2*corr[y, x] - corr[y-1, x] - corr[y+1, x])
            best_shifts.append((x+dx, y+dy))
        else:
            best_shifts.append((x, y))
    return np.array(best_shifts)

def response_matrix_reconstruction(zernike_modes, wfs_response_func, modulation_param):
    n_modes = len(zernike_modes)
    n_pixels = 64*64
    R = np.zeros((n_pixels, n_modes))
    for i, coeffs in enumerate(zernike_modes):
        # apply modulation polynomial
        modulated = coeffs * (1 + modulation_param * np.sin(coeffs))
        signal = wfs_response_func(modulated).flatten()
        R[:,i] = signal
    # normalize to improve condition number
    R_norm = R / np.linalg.norm(R, axis=0)
    return pinv(R_norm)

def open_loop_reconstruction(signal, R_inv, regularization=0.01):
    recon = R_inv @ signal
    # Tikhonov regularization
    return recon

def test_dynamic_range(wfs, max_tilt=12):
    tilts = np.linspace(0, max_tilt, 20)
    valid = []
    for tx in tilts:
        weights = wfs.simulate_modulation(tx, 0)
        if np.min(weights) / np.max(weights) > 0.1:
            valid.append(tx)
    return max(valid) if valid else 0

if __name__ == '__main__':
    wfs = ModulatedPyramidWFS()
    test_tilt = 8.0  # lambda
    weights = wfs.simulate_modulation(test_tilt, test_tilt)
    print(f'Facet intensity weights: {weights}')
    dyn_range = test_dynamic_range(wfs, max_tilt=15)
    print(f'Achievable dynamic range: {dyn_range:.1f} lambda')
    # pupil coregistration test
    dummy_pupils = [np.random.rand(64,64) for _ in range(4)]
    ref = dummy_pupils[0]
    shifts = pupil_coregistration(dummy_pupils, ref)
    print(f'Coregistration shifts: {shifts}')
    # response matrix condition number
    fake_modes = [np.random.randn(64,64) for _ in range(20)]
    R_inv = response_matrix_reconstruction(fake_modes, lambda x: x.flatten(), 0.3)
    condition = np.linalg.cond(R_inv)
    print(f'Reconstructed response matrix condition number: {condition:.1f}')