A Bus Passenger Flow Statistics System Based On 3D Vision

In recent years,the rapid development of public transport,bus has become a preferred way for residents to travel,but at present there are many problems in urban public transport.Because of the huge traffic flow of urban buses,but there are many unreasonable places in bus scheduling,Especially during the morning and evening rush hour,the unreasonable planning of bus lines,resulting in unreasonable arrangements for a certain line or too few passengers in a car and other issues.In order to solve these problems,an accurate real-time passenger flow data will become a key issue.At present,due to the continuous launch of smart city construction,passenger flow data has become particularly important,and passenger flow statistics have become more and more important.An indispensable part of the construction of smart cities,how to obtain accurate and real-time passenger flow data has become a research hotspot today.After obtaining accurate and real-time passenger flow data,unreasonable routes can be planned based on the passenger flow data,which is more scientific The public transportation dispatching can solve the various problems that have appeared at this stage and make it easier for residents to travel.In the current era of big data,accurate and real-time passenger flow data can also bring many good economic benefits to the city.This article mainly describes the method of video statistics for passenger flow data statistics,using a structured light-based depth camera as the passenger flow data collection device for statistics,using the depth camera as the collection device,installed directly above the front and rear doors of the bus for passenger flow data Collection.This article first collects the video data by the depth camera,preprocesses the obtained video information,uses binarization to process the image,and proposes an improved median filter algorithm to denoise the image and remove the influence of the background on the target information.,To reduce the difficulty of passenger target recognition.In order to further reduce the influence of background and other factors on target detection,a hierarchical depth map denoising algorithm based on wavelet threshold is proposed.The algorithm estimates the noise intensity of the image based on the acquired depth image,then calculates the interval between the depth layer levels,layer the image,selects the layer that needs to be denoised,performs wavelet threshold denoising,and divides the denoising completed The layer images are finally integrated into a complete depth map.Through this algorithm,many difficult-to-remove background noises or noises caused by other factors are used to further improve the accuracy of target detection.Finally,the inter-frame difference method is used to detect moving targets.Because the body and background are static when passengers get on and off the car,this method can detect the target's head features well.After detecting the target,the target's head feature is proposed.The centroid represents the target for tracking,using the principle of the shortest centroid distance of the same target for multi-target tracking,and it is proposed to set a double counting line in the tracking frame to count the passenger flow data.In this dissertation,the passenger flow system uses the Firefly development board,the software development platform is VS2017,using some Open CV functions and their own image processing algorithm to process the collected passenger flow data,Then the collected video data are processed to obtain the final count.The system is tested on the bus.The results show that the system can achieve accurate and real-time statistics of passenger flow data,and can meet the requirements of the bus company.