| At present,countries around the world are paying more and more attention to the construction of smart cities,and urban management systems are a key step in the construction of smart cities.In China,some cities have applied urban management systems to urban management,but they rely on special personnel to identify,distribute and handle cases.This makes the system inefficient and wastes a lot of financial and material resources.Therefore,in view of this problem,this paper deeply studies the video image target detection algorithm of urban management cases,and further integrates the target detection algorithm into the urban management system to automatically identify the case types,improve the intelligence level of the urban management system,and promote urban management.Efficient.The specific research contents are as follows:1)A vehicle violation detection algorithm is proposed to realize the detection of illegal and illegal cases of motor vehicles and non-motor vehicles,which improves the stability of such cases.The method uses homomorphic filtering to preprocess the samples to eliminate the influence of environmental factors such as illumination on the recognition results.Then,the ZCA whitening process is applied to reduce the correlation between the data,and the improved convolutional neural network is used to extract the sample features.The extracted sample features are sent to the detection network for detection;finally,tracking and motion analysis are used to determine illegal vehicles in the region of interest(ROI).The effectiveness of the method is verified by comparison with Faster R-CNN,SSD and YOLO.Experimental results show that the method has stronger robustness and higher detection accuracy.2)An urban management case identification algorithm based on improved convolutional neural network is proposed,which is aimed at scribbling,non-motor vehicle parking,shared bicycle parking,exposed garbage,traffic barrier damage,and rainwater operator damage.Eight kinds of urban management cases,such as motor vehicle arbitrage and damage to the manhole cover,were applied and verified,and compared with the traditional methods,the detection accuracy and detection efficiency were improved.The algorithm first uses the MSRCR algorithm to preprocess the samples;then introduces a 1×1 convolution in the linear convolutional layer of the detection network,which enables the network to extract samples to more abstract nonlinear features;As the evaluation network,the network scores the relevant information of the main objects extracted by the detection layer to obtain the recognition result of the case.The Fast R-CNN,Faster R-CNN,YOLO and SSD algorithms are used as the comparison algorithm.The experimental results show that the proposed algorithm has the best robustness and recognition accuracy.And in the VOC2007+2012 dataset,compared with six algorithms such as 100 Hz DPM,30 Hz DPM,Fast R-CNN,Faster R-CNN,YOLO and SSD,and achieved good experimental results,fully illustrating the improved convolution.The effectiveness of neural network algorithms.In summary,this paper integrates the target detection algorithm based on deep learning into urban management,which is indispensable for the improvement of the efficiency of urban management processing cases,the saving of labor costs such as human resources and the construction of smart cities. |
