The exigency for precise intelligent detection and recognition technology aimed at infrared targets within intricate background environments is pressing, bearing crucial practical implications for the evolution of advanced reconnaissance equipment and guided weaponry. Within this context, the present article introduces a refined cascaded neural network model as a promising solution. Initially, a context information enhancement structure is seamlessly integrated at the culmination of the recommendation network, strategically amplifying the performance of the upgraded cascade method dedicated to infrared target detection. Subsequently, the model leverages a multi-task separation branch attention mechanism to engender independent feature channel attention for distinct subtasks. This strategic approach effectively mitigates performance degradation stemming from undue information sharing among subtasks, thereby elevating the accuracy of target recognition. Comprehensive training and recognition comparison experiments, conducted employing the same dataset utilized in other target recognition algorithms, serve to affirm the exceptional recognition accuracy achieved by this neural network model. This validation underscores the efficacy and potential practical utility of the proposed enhancements, positioning this model as a significant stride forward in the domain of target recognition technologies.