Young learners' vocabulary learning needs interaction with language input when they are engaged in an activity. Given that AI-supported image recognition technologies offer hands-on learning in authentic contexts, and that self-regulated learning (SRL) enables learners to monitor and evaluate their learning when interacting with multi-sensory experiences of the target vocabulary, this study investigated how AI image recognition technologies along with SRL affect students' vocabulary acquisition, self-regulation, and learning anxiety in the language classroom, and further analysed students' learning behaviours during learning. A total of 47 Grade 3 students participated. Students in the experimental group (EG) assigned to AI-supported image recognition (AI-IR) with SRL, learnt with vocabulary from realia using AI-IR with SRL. Students in the control group (CG) assigned to AI-supported optical character recognition (AI-OCR) technology with SRL, learnt with vocabulary from printed texts using AI-OCR with SRL. Results indicated that the EG students significantly outperformed the CG students in terms of vocabulary acquisition, although no significant differences were found in the two groups' self-regulation or learning anxieties. Behavioural analysis revealed that the EG students had greater development of their self-regulation and lower learning anxiety. Lastly, suggestions and limitations are provided for further research.