基于改进的最大类间方差法的水果图像识别研究

Research on Fruit Image Recognition Based on Improved Maximum Between-Class Variance Method

针对水果自动识别过程中特征信息提取不完整的问题,本文提出一种基于改进的最大类间方差法OTSU对水果图像进行分割。通过对图像中值滤波处理降低随机噪声的干扰,增大目标图像和背景之间像素值与最佳分割阈值之间的差值,使目标图像与背景图像与各自类间中心的距离尽可能相近,达到相对方差取代绝对方差实现图像分割,然后对目标图像提取颜色特征和形状特征实现不同种类的水果图像识别。实验结果表明,改进后的OTSU所得阈值能分割到更加清晰的图像,图像分割的运行时间明显缩短,水果图像识别的平均正确识别率提高了15%左右。该研究提高了水果识别的效率,具有一定的实际应用价值。

In view of solving the problem of incomplete feature information extraction in fruits recognition, this paper proposes an improved OTSU method to segment fruit images. First, the image processing median filtering is used to reduce the interference of random noise, and then increase the difference between target and background pixels and the best segmentation threshold, at the same time let the target image and background image with the two respective distance between center as close as possible, and finally achieve relative variance instead of absolute variance to achieve image segmentation. After image segmentation, color features and shape features of target images are extracted to complete image recognition of different kinds of fruits. The experimental results show that the threshold obtained by the improved OTSU can obtain clearer segmentation images. Meanwhile, the running time of the improved OTSU algorithm for image segmentation is significantly shortened, and the average correct recognition rate of fruit image recognition is increased by about 15%. This study improves the efficiency of fruit identification and has certain practical application value.