面向苹果品质检测的图像融合方法

Image Fusion for Apple Quality Detection

摘要:

水果的损伤不但影响外观，还会造成营养流失和病菌感染等问题。作者提出一种面向水果品质检测的图像融合方法，为果农或经销商提供水果损伤信息。首先，对苹果的可见光图像和红外图像分别使用四元数小波变换得到低频子带与高频子带；接着，对低频子带使用基于局部平均梯度的加权平均方法进行融合，高频子带采用基于导向滤波的绝对值取大的融合规则实现融合；最后，将融合后的高频系数与低频系数使用逆四元数小波变换得到融合后图像。实验结果表明，提出的方法很好的保留了可见光图像中表皮的细节纹理和红外图像中检测到的次表面损伤信息。该方法能够提高苹果品质检测的准确性。

Abstract:

The damage of fruit not only affects the appearance quality，but also causes nutritional loss and bacterial infection. The purpose of this paper is to research an image fusion method for quality detection of fruit and then provide the damage information of fruit to farmers or distributors，which help them take early means to avoid greater economic losses. Firstly，low-frequency and high-frequency subbands are obtained by implementing the quaternion wavelet transform（QWT） on the visible image and the infrared image of apple. Secondly，the low-frequency coefficients are fused using the weighted average fusion rule based on the local average gradient. The high frequency coefficients are integrated by choose-max fusion rule based on guided filter. Finally，the fused image is obtained by using the inverse QWT after the fusion of the low frequency coefficients and high frequency coefficients. The experimental results show that the proposed method can preserve the detail texture of the epidermis in the visible light image and the secondary surface damage information detected in the infrared image. Therefore，the proposed method is helpful for improving the quality detection accuracy of apples.