Traffic sign dynamic tracking detection optimization method based on Nest DNN algorithm

Abstract

The invention relates to a traffic sign dynamic tracking detection optimization method based on a Nest DNN algorithm, and the method comprises a step of constructing a resource perception scheduling scheme, and the step of constructing the resource perception scheduling scheme comprises the steps: setting a cost function C for vehicle-mounted equipment according to a formula, so as to guarantee the processing precision and processing speed of a derived model mv; scheduling scheme design is carried out on a detection model and an identification model which run at the same time, minimization optimization is carried out on a program with the maximum cost, a resource perception scheduler is optimized through constraint conditions to allocate running resources to all concurrent application programs fairly, and therefore the performance of the concurrent application programs is balanced. The method has the advantages that the defects in the prior art are overcome, the resource use efficiencyof the vehicle-mounted mobile terminal equipment is improved, the model operation efficiency is improved, the energy consumption is reduced, the traffic sign detection and recognition effect of the vehicle in the driving process is improved. And the safety and standardization of vehicle operation can be effectively improved in an automatic driving popularization stage in the future.

Description

technical field [0001] The invention relates to the technical field of advanced driving assistance systems in automatic driving, in particular to a method for optimizing dynamic tracking detection of traffic signs based on NestDNN algorithm. Background technique [0002] In recent years, the number of motor vehicles in my country has been increasing. According to the statistics of the Ministry of Public Security, as of June 2020, the number of cars in the country has reached 270 million. The traffic accidents caused by this have brought huge losses of life and property every year. [0003] Faced with such problems, advanced driving assistance systems and even automatic driving have become one of the popular solutions in recent years. As one of the key technologies of advanced driving assistance systems, traffic sign detection and recognition can quickly detect traffic signs on the road and Accurate identification of signs can broaden the driver's field of vision, transmit ro...

AI Technical Summary

Problems solved by technology

[0004] However, due to the resource requirements of the deep learning model, multiple parallel applications will compete for resources when resources are limited, resulting in a low frame rate for streaming video processing; on the other hand, when there are additional runtime resources, Compression models also cannot take advantage of these extra resources to fix accuracy loss

Although the detection and recognition of traffic signs are two sequential programs running linearly, due to the continuity of road detection, it is still necessary to ensure the detection process of traffic signs in the process of traffic sign recognition. The intelligent processing and analysis system cannot meet the existing deep learning model resource requirements, and there will be problems in the accuracy of detection and recognition. If more resource requirements are sought from the cloud, additional delays in timeliness will occur

Moreover, due to the excessive resource requirements of the deep learning model, the current advanced intelligent assistance system often leads to problems such as resource occupation, video freeze, low recognition accuracy, or slow processing speed, which cannot meet the requirements of engineering applications.

Images