Research And Implementation Of Vehicle Hub Recognition And Classification Algorithms

With the improvement of people's living standards and the development of science and technology,the rise of China's automobile industry has been promoted.Therefore,the traditional production model of automobile Hub hub has not kept up with the development speed of today's market.Due to the limited space in the hub production workshop,a hub transmission line is usually used to transport multiple types of hubs at the same time.In the traditional automobile Hub hub production process,the hubs are sorted manually,which is low in manual operation efficiency,and easy to cause visual fatigue and inattention after a long working time.In order to solve this problem,in recent years,the automatic technology of hub sorting has been developed and applied.In order to improve the recognition accuracy of the hub identification and classification technology,this paper has made further research on the hub identification and classification algorithm.In this paper,the particle swarm optimization algorithm is improved to optimize BP neural network and convolutional neural network respectively,and then construct two kinds of Hub recognition and classification algorithms.The main research work and results are as follows:(1)In order to improve the accuracy of the hub identification and classification,analyze the usual noise types and noise sources,and combine the classical filtering technique to achieve the Hub image pre-processing operation.Firstly,the grayscale image is filtered.In this paper,the median filtering technique is adopted to protect the hub feature information of this paper while achieving the purpose of denoising.Secondly,the dynamic threshold is selected by genetic algorithm to realize image segmentation.The morphological processing method is used to further denoise;finally,the Canny edge detection algorithm is used to extract the image edge information.The purpose of image preprocessing is to obtain better result of the hub feature extraction and improve the accuracy of feature data.(2)The key technique after image preprocessing is feature extraction.In this paper,six features are extracted according to the hub feature information as the data dependent in the later pattern recognition,which are the outer contour of hub,the center small circle,the width of large hole,the number of spokes,the number of small holes,and the distance between the large hole and the center of the circle.In this paper,the improved Hough algorithm and statistical principle are used to study the extraction methods of each feature,and the effectiveness of the above-mentioned hub feature extraction method is evaluated on Matlab platform.(3)Based on hub feature extraction,this paper proposes a hub recognition and classification algorithm based on improved particle swarm optimization the BP neural network.BP neural network has better fault tolerance,but it is not easy to converge during network training.Therefore,particle swarm optimization is proposed to optimize BP neural network.Firstly,aiming at the defects of standard particle swarm,such as prematurely maturity,the improvement of particle swarm is realized by changing the inertia weight of particle swarm,introducing genetic factors,speed limit,and other algorithms.Secondly,the improved particle swarm optimization BP neural network is used to improve the convergence speed and accuracy of network training,and the algorithm is used to realize the recognition and classification of Hub hubs.Then the improved particle swarm optimization algorithm is compared with the particle swarm optimization algorithm of others to verify the effectiveness of the algorithm.Finally,the algorithm is platform-based.And,a practical software for Hub identification and classification is designed.(4)In order to avoid the incomplete extraction information of the specified hub features,the BP neural network recognition rate may be unsatisfactory about similar hub,this paper proposes a hub recognition and classification algorithm based on optimized convolutional neural network.The specific method is to optimize the weight parameters by using the improved particle swarm optimization algorithm and the maximum-median pooling method in this paper.In order to modified PSO and the pooling method,the experimental results are compared with other related algorithms,which proves that the algorithm is effective.