基于AlexNet的高精度图像亮度计测试方法

    Test method of high precision image luminance meter based on AlexNet

    • 摘要: 图像亮度计能够高效检测LED显示光源的亮度、色度及空间光分布,但在测试单色LED时,因其光谱带宽较窄,采用标准A光源标定的传统图像亮度计存在光谱失配问题,导致测试误差较大。针对这一问题,以LED光源输出的多种光谱为基础,开展图像亮度计测试方法研究。首先,分析不同LED颜色对测试结果的影响,基于AlexNet模型构建改进的卷积神经网络,实现光源颜色识别,在此基础上建立亮度与色度的数学模型,用于校正并计算光源的亮度和色度;其次,研究图像亮度计输出的灰度信号与相机曝光时间之间的数学关系,结果表明二者呈线性关系,这为测试校正提供了重要依据。与传统标准A光源标定方法不同,所提出的方法利用LED光源输出的多种光谱进行测试,最终将LED亮度与色度的测试误差控制在5%以内,为图像亮度计测量精度的提升提供了一种有效解决方案。

       

      Abstract: Image luminance meter can effectively detect luminance, chroma and spatial light distribution of LED display light source. However, when testing monochrome LED, due to its narrow spectral bandwidth, traditional image luminance meter calibrated with standard A light source has spectral mismatch problem, which leads to large test error. In order to solve this problem, this paper studied the test method of image luminance meter based on output spectra of LED light sources. Firstly, the influence of different LED colors on test results was analyzed, and an improved convolutional neural network was constructed based on Alexnet model to realize light source color recognition; on this basis, the mathematical model of luminance and chroma was established to correct and calculate luminance and chroma of light source. Secondly, the mathematical relationship between grayscale signal output by image luminance meter and exposure time of camera was studied. Results indicate that the two are linearly related, which provides an important basis for test calibration. Different from traditional standard A light source calibration method, this method uses multiple spectra output by LED light sources for testing, and finally controled test error of LED luminance and chroma within 5%, which provides an effective solution for improving measurement accuracy of image luminance meters.

       

    /

    返回文章
    返回