At present, the evaluation of cigarette retail terminals mainly adopts field data collection and human evaluation. Due to the large number of terminals, complex evaluation criteria and heavy evaluation workload, there are such problems as low evaluation efficiency, limited coverage, strong subjectivity and unreliable results. To solve the problems above, a cigarette retail terminal evaluation system based on image recognition is designed and developed, which supports the image acquisition, analysis, evaluation and visualization of retail terminals. Image acquisition by multiple characters and image quality inspection ensures the convenience, coverage and quality of image acquisition, guarantees the objectivity, accuracy, real-time and reliability of terminal evaluation, and greatly reduces the workload of customer managers in terminal evaluation. The average precision of terminal image optical character recognition is 95.4%. The average precision and correct recognition rate of cigarette box and carton recognition are 91.3% and 98.6%, respectively. The system has good robustness for tricky cases such as large angles and folded image recognition. It can effectively support cigarette retail terminal evaluation, provide efficient and high-quality data support for tobacco business enterprises to guide retail clients' operations and formulate precision marketing strategies and other business scenarios, and promote the digital transformation of the team.