Improved Kernel Filtering Algorithm Research And Its Application In Smart Kitchens

With the development of society and the progress of science and technology,the issue of food safety has increasingly attracted people’s attention and become a topic of increasing concern.Traditional food safety monitoring relies on manual monitoring,and its accuracy is not high.The target recognition and tracking technology can automatically realize the positioning and tracking of the target.Applying it to the kitchen environment can realize the real-time monitoring of the moving target in the food processing process and improve the efficiency of food safety monitoring.At the same time,the technology can also monitor and manage the equipment and personnel in the kitchen,detect the abnormal situation in time and take corresponding measures immediately to ensure the safety and health of the kitchen.KCF algorithm is a target tracking algorithm based on kernel correlation filtering,which uses the samples of training set and test set for kernel regression to obtain the estimated position of the target.The algorithm performs well in terms of running speed,tracking accuracy and tolerance to target change,but the effect is not satisfactory in small target tracking and occlusion tracking.In the process of small target tracking,there are problems such as insufficient target feature extraction and low sampling rate,which lead to inaccurate target recognition.Scale pool algorithm can effectively improve the performance of small target feature extraction,size change and sampling rate,and is widely used in small target detection.This paper proposes a tracking method that integrates the scale pool algorithm into the KCF algorithm,strengthens the feature extraction and sampling rate of small targets,and increases the accuracy of tracking by updating the template.In the kitchen environment,there are a lot of equipment to block the target,and the target will disappear during the tracking process.Byte Track algorithm is a target tracking algorithm based on deep learning,which is very suitable for target tracking under occlusion,but the accuracy of this algorithm depends on the target detection results.This paper proposes a joint tracking method of improved KCF algorithm and Byte Track algorithm.The improved KCF algorithm is used to obtain better detection results,and the combined Byte Track algorithm is used to process occluded targets,making full use of local information and global characteristics of the target,and improving the accuracy and robustness of tracking.Based on the above two algorithms,this paper implements an intelligent monitoring system for smart kitchen.The system is divided into two parts: hardware and software.The main control chip of the hardware part is STM32,which has the functions of real-time acquisition of activity targets,kitchen environmental parameters,water and electrical resources parameters,information transmission,online monitoring,early warning,database storage,etc.The software part is mainly implemented through Spring+Mybatis+spring Boot,with Redis caching mechanism.The display layer architecture design is implemented through Vue+Element+Echart and deployed on NGINX.The improved algorithm and system are tested through the actual built experimental environment.The results show that the improved KCF algorithm,the improved KCF algorithm and the combined tracking algorithm of Byte Track algorithm proposed in this paper have the advantages of high recognition rate,accurate positioning and fast tracking speed for small and occluded targets in the actual kitchen environment.It is applied to the smart kitchen monitoring system,with the characteristics of low computation and small network deployment delay,It can meet the needs of practical applications.