Research And Implementation Of Image Recognition Algorithm In Mobile Edge Computing Environment

With the increasing application of image recognition in the mobile phone,the timeliness of recognition and storage requirements is getting higher and higher.The processing capability and storage capacity of the mobile terminal device cannot meet the requirements,and the energy consumption of the terminal device cannot be ignored in the data transmission process.Therefore,the computation offloading is required.The existing solution is to offload computing tasks to the cloud server,but this may cause a lot of problems,such as excessive delay.On the other hand,it is important to extract key features for image recognition.The existing feature extraction techniques,including Linear Discriminant Analysis,Marginal Fisher Analysis,Discriminant Neighborhood Embedding,etc.,take into account of the optimization of intra-class scatter and inter-class scatter simultaneously,but the trouble is that the large discrepancy between the sum of intra-class distances and the sum of inter-class distances in different datasets leads to an overwhelming impact on the intra-class distance,thus resulting in some important information loss.Then the solutions to the above problems are proposed in this paper.The main innovations are:An optimization algorithm for cognitive service feature extraction is proposed,and it is based on deep learning layered extraction of effective features.First of all,the homogeneous data is aggregated by minimizing the sum of the distances of the intra-class.Secondly,the heterogeneous data is separated through maximizing the sum of the distances of the inter-class.fore,the extracted features are more effective in identifying cognitive services.Finally,the optimization algorithm is compared with other classical feature extraction algorithms,and the experimental results indicate that the optimization algorithm can effectively improve the recognition accuracy.The proposed image recognition optimization algorithm is implemented and deployed in the mobile edge computing environment.In this application,an APP for acquiring images and receiving image recognition results is implemented in the mobile terminal.Then the algorithm is deployed on the remote cloud to deliver the feature information database and the projection matrix to the edge server.Finally,the edge server obtains the feature information of the uploaded image in the mobile terminal thr-ough the projection matrix,compares it with the feature information database,and then returns the recognition results to the mobile terminal.The experimental results indicate that the image recosnition based on the mobile edge computing architecture can effectively reduce both network transmission and response time while ensuring the recognition rate.