Design and Performance Evaluation of an Intelligent Clothing Quality Inspection System Combining Artificial Intelligence and the Internet of Things

In the process of clothing production, changes in environmental factors such as lighting, temperature, and humidity can affect image quality and detection accuracy, but existing research often does not fully consider these factors. Therefore, this article combines the image recognition technology of artificial intelligence algorithms and the Internet of Things to design a clothing quality inspection system. Firstly, a clothing image preprocessing method based on smooth filtering is studied, and an improved CBAM (Convolutional Block Attention Module) mechanism is introduced into the Convolutional Neural Network (CNN) algorithm to adjust the attention of the algorithm's information channels and regions. Secondly, the Laplacian variance method is used to measure the clarity of image samples, and multiple clothing samples from the DeepFashion database are selected for algorithm training and quality inspection experiments. The experimental results show that the introduction of channel attention mechanism (CA) significantly improves the accuracy of the model, reaching between 93.8% and 95.2%. In addition, the test results showed that wrinkles accounted for 37% of clothing defects and thread ends accounted for 30%, indicating that wrinkles and thread ends are the most common types of defects. Taking T-shirts as an example, the accuracy of artificial intelligence detection is 95%, which is higher than the 93% of manual detection, an increase of 2 percentage points.