改进DHCEP模型在图像智能识别中的应用研究

Application of improved dhcep model in image intelligent recognition

针对现有图像识别方法在识别目标较多时存在的准确率低和效率低等问题。基于图像识别系统,提出了一种结合DHCEP选择性集成模型和改进YOLO v3模型的中餐图像识别方法。改进后的YOLO v3模型用于检测中餐的位置,并在图像中保留中餐部分。DHCEP集成了AlexNet、SqueezeNet和GoogLeNet模型来识别图像中的中餐类型。通过对单一模型与常规选择性集成模型的对比分析,验证了该方法的优越性。结果表明,在目标检测中,改进后的YOLO v3模型的检测准确率达到了97.09%,在目标识别中,DHCEP集成模型投票连接法的最高准确率为89.56%,平均概率连接法准确率为90.15%。该图像识别方法为识别技术的进一步发展提供助力。

Aiming at the problems of low accuracy and low efficiency in the existing image recognition methods when there are many recognition targets.Based on the image recognition system,a Chinese food image recognition method combining dhcep selective integration model and improved Yolo V3 model is proposed.The improved Yolo V3 model is used to detect the position of Chinese food and retain the Chinese food part in the image.Dhcep integrates alexnet,squeezenet and googlenet models to identify Chinese food types in images.The superiority of this method is verified by comparing the single model with the conventional selective integration model.The results show that in target detection,the detection accuracy of the improved Yolo V3 model reaches 97.09%,in target recognition,the highest accuracy of the dhcep integrated model voting connection method is 89.56%,and the accuracy of the average probability connection method is 90.15%.The image recognition method provides help for the further development of recognition technology.