The existing portable crude oil water content measurement devices and online water content measurement devices have poor accuracy and high installation costs. The production status of oil fields still relies mainly on manual sampling and detection methods for crude oil water content measurement. This paper proposes a crude oil water content analysis method based on image recognition technology, which intelligently identifies the oil-water interface through image acquisition, segmentation, and image grayscale recognition of the oil-water separated sample. The designed pixel ratio algorithm intelligently identifies the water content of the oil product, and the effectiveness of the algorithm is verified through sampling analysis. A portable device is designed and developed to precisely measure the wellhead water content. Compared with the manual water content detection, a crude oil water content measurement is designed for this device. The intermittent sampling mode is adopted to automatically realize some functions, such as heating, stirring, precipitation, data reading, and data transmission. On-site application test confirmed the reliability of this device’s detection under the various scenarios.