Guangzhou Nanyang Polytechnic College, Conghua 510925, China.
Guangzhou Nanfang College, Conghua 510970, China.
Comput Intell Neurosci. 2022 Aug 23;2022:2926669. doi: 10.1155/2022/2926669. eCollection 2022.
With the development of the new generation of technological revolution, the manufacturing industry has entered the era of intelligent manufacturing, and people have higher and higher requirements for the technology, industry, and application of product manufacturing. At present, some factories have introduced intelligent image recognition technology into the production process in order to meet the needs of customers' personalized customization. However, the current image recognition technology has limited capabilities. When faced with many special customized products or complex types of small batch products in the market, it is still impossible to perfectly analyze the product requirements and put them into production. Therefore, this paper conducts in-depth research on the improved model of product classification feature extraction and recognition based on intelligent image recognition: 3D modeling of the target product is carried out, and various data of the model are analyzed and recorded to facilitate subsequent work. Use the tools and the established 3D model tosimulate the parameters of the product in the real scene, and record them. Atthe same time, various methods such as image detection and edge analysis areused to maximize the accuracy of the obtained parameters, and variousalgorithms are used for cross-validation to obtain the correct rate of the obtaineddata, and the standard is 90% and above. Build a data platform, compare simulated data with display data by software and algorithm, and check by cloud computing force, so that the model data can be as close to the parameters of the real product as possible. Experimental results show that the algorithm has high accuracy and can meet the requirements of different classification prospects in actual production.
随着新一代技术革命的发展，制造业已经进入智能制造时代，人们对产品制造的技术、产业和应用提出了更高的要求。目前，一些工厂已经将智能图像识别技术引入生产过程，以满足客户个性化定制的需求。然而，目前的图像识别技术能力有限。当面对市场上许多特殊定制产品或复杂类型的小批量产品时，仍然不可能完美地分析产品需求并将其投入生产。因此，本文对基于智能图像识别的产品分类特征提取和识别改进模型进行了深入研究：对目标产品进行 3D 建模，并分析和记录模型的各种数据，以便于后续工作。使用工具和建立的 3D 模型模拟真实场景中的产品参数，并记录下来。同时，采用图像检测和边缘分析等各种方法，最大限度地提高所获得参数的准确性，并对各种算法进行交叉验证，以获得所获得数据的准确率，准确率标准为 90%以上。建立数据平台，通过软件和算法比较模拟数据和显示数据，并通过云计算力进行检查，使模型数据尽可能接近实际产品的参数。实验结果表明，该算法具有较高的准确性，能够满足实际生产中不同分类前景的要求。
