Research on Internet of Things Image Compression Method Based on Convolutional Self Coding Fusion

Objective In order to solve the problem that the cloud platform occupies too much data storage space of the Internet of Things in the fields of intelligent manufacturing, intelligent medical treatment and intelligent automobile, and the amount of image data is too large, this paper proposes an Internet of Things image compression method based on convolutional self-coding fusion, and researches the compression method of the collected Internet of Things images. Method Firstly, the collected Internet of things images were grayed to obtain grayscale images. Secondly, the improved convolutional autoencoder (GCNN) model was used to encode, binarize and decode the original color image. Finally, the image decoded by the autoencoder and the gray image were fused and enhanced by the DHSV fusion reconstruction method. Result The proposed method uses the BSDS500 dataset, in which the training set consists of 200 images, the test set consists of 200 images, and the validation set consists of 100 images. The network structure was built according to the way in Section 2.2, the learning rate was set to 0.01 and 0.1 respectively, and the training was performed 100 times with 10 iterations per epoch. Finally, Set8(8 images) is used for testing. The color image is used to evaluate the reconstruction results in visual effect, and the Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) are used to evaluate the reconstruction results. The experimental results show that the compression ratio can reach 2.28, and the compressed space is reduced by nearly 1/2. At the same time, the fusion reconstruction image can reduce the distortion, and maintain a good level of clarity, contrast and information richness. Conclusion The network model and fusion reconstruction method proposed in this paper can quickly extract image features while reducing network parameters, and can fuse the high-resolution information of grayscale images to achieve good reconstruction results, which basically meets the requirements of cloud transmission and storage in the Internet of things scene, and can provide reference for later research.