The use of computer vision to count the number of people in the elevator car is a new topic in the field of elevator safety research.At present,most of computer vision detection algorithms are implemented by using PC platform,and their practicability is poor.However,how to extend the computer vision detection algorithm to the implementation of the embedded platform is a diffi-cult research point.Therefore,it is of great research value and significance to study the number of personnel in the elevator car based on embedded DSP plat-form and using computer vision.This article first elaborates several methods of using computer vision to identify the number of people in the current general commercial elevator envi-ronment.After analyzing and comparing with the background of construction environment,AdaBoost algorithm was chosen to identify the number of con-struction workers wearing helmets,so as to achieve the indirect statistics of the number of elevator personnel.After that,a set of engineering elevator statistics system based on C674x-DSP as the core processing module was designed.The detailed design flow of the system platform is given,including the design of the hardware structure of the video acquisition module,the conversion of the signal sampling format and the flow of work,and the implementation of the software and hardware of the network communication module.Next,with the OpenCV computer vision library and the VS2013 software on the PC platform,the collected positive and negative sample sets were trained to obtain the helmet detection classifier and verify the accuracy of the detection on the PC.Then,the key OpenCV detection function is recompiled under the CCS software environment,so that the classifier detection program can run correctly on the embedded DSP platform.Finally,using DSP platform optimization techniques,Cache optimization,software pipelining,optimization of inline functions and other optimization methods to further reduce the detec-tion time.Therefore,the C674x-DSP engineering elevator number statistics system designed in this paper has practical value in engineering while ensuring accuracy. |