Research On Computer Vision Analysis Method Of Potential Customers For Automobile Sales

In the automotive sales industry,customer resources are an important part of sales performance.How to tap potential customers and how to collect basic information about potential target customers is the key to car sales.With the rapid development of computer vision technology,face detection,face tracking,and face recognition technologies are widely used in various fields such as access control,surveillance,and smart stores.Under the premise of no human interference,how to use computer vision methods to obtain customer intent frequency,stay time,stay area and other purchase intent information,so as to provide accurate product introduction and services to customers,to achieve the identification and analysis of potential customers,this pair It is of great significance to establish potential customer consumption information and achieve precise marketing.This article takes the actual project of Changan Automobile 4s shop as the research background,and uses computer vision analysis methods to mine the basic information of potential customers as the demand-oriented.It mainly includes the following aspects:(1)Face detection: based on skin color model and improvement AdaBoost algorithm combined with face detection method.This method converts the image from RGB color space to YCbCr color space with better skin color clustering characteristics and distribution rules by determining whether the collected face image needs illumination compensation,and establishes a Gaussian model and performs morphological processing and shape filtering on the image To obtain a candidate area containing a human face.Then,the obtained face candidate region is subjected to face detection by the improved AdaBoost algorithm,and the final face detection result is obtained.Experimental results show that this method can effectively reduce the rate of face misdetection and improve the accuracy of face detection;(2)Face tracking: A face tracking algorithm based on LBP feature-based TLD algorithm and kalman filter is proposed.The algorithm extracts the grayscale invariance LBP characteristics of the face to enhance the accuracy and robustness of the subsequent TLD target tracking algorithm to adapt to the face tracking under the lighting changescene,and aims at the local and short-term face of the face in practical applications.Full occlusion is easy to cause the problem of face tracking failure.A partial and short-term full occlusion scene prediction and tracking method based on Kalman filtering is proposed.The comparative experimental results show that the face targettracking algorithm based on the LBP feature-based TLD algorithm combined with the Kalman filter can better achieve accurate tracking of the customer's face in the 4S store,and it can be fully blocked in the light changes,partial or short-term face In the scene,the algorithm has good robustness.According to the established face tracking algorithm,it can realize multi-frame extraction of the face image of the customer after entering the store,and record the time of entering and leaving the store and the stay time of the area of interest,which establishes dimensional information for the subsequent potential customer mining.(3)Label-free identification: A label-free identification method based on PCA + LDA is proposed.First,the face image is subjected to histogram equalization and face geometry normalization processing to eliminate the interference of lighting,posture and background,so that the image is standardized.Secondly,the PCA + LDA algorithm is used to extract the features of the face image,and finally the SVM algorithm is used for classification and recognition.The experimental results show that the proposed standard-free identification algorithm has good accuracy;(4)The potential customer identification systemis developed.Based on the research of face detection,face tracking,and label-free identity recognition algorithms,based on sales dimension information such as the frequency of potential customers visiting the store,time of stay,area of interest,etc.,using the QT platform,a potential customer mining system for car sales was developed.The system can better extract a series of face images of customers entering the store,automatically record the time information,time of stay and active area ofthe store,and based on the historical unlabeled face image library,determine whether its new customer is old Customers,thus providing an effective computer vision analysis method for mining potential customers for automobile sales.