Event detection is the remarkable technology through which the search, review, and archiving the events takes place. The event detection becomes more complex when the quality of the image extraction becomes poor. Further, the detection of events based on the semantic cues of the provided image remains more critical as they resemble each other. Even though several researches exist, the researches failed to detect the most accurate event detection based on the cues, which resulted in poor performance. To enhance the detection accuracy, the Border Collie optimization-based Convolutional Neural Network classifier (BCO-based CNN) is proposed in this research. The efficacy of the research is enhanced with some feature extraction methods such as Hybrid Ternary Pattern (HTP) and Scale-Invariant Feature Transform (SIFT). The extraction of these features aids the model in achieving enhanced detection of complex events. The performance of complex event detection based on the BCO-based CNN method provides better comparison results with existing methods reported as 88.773% accuracy, 89.299% sensitivity, and 89.821% specificity for the USED dataset.