Motifs are not stranded, neither is transcription factor binding a strand-specific event. They bind double stranded DNA recognizing a motif. If a sequence is overrepresented in your selection it will be overrepresented no matter if providing the top strand or its reverse complement. So no, you can use intersect as usual, only the coordinates matter. The only thing I would do though is to limit the peak to like 200bp from the peak center. THis is what motif tools such as Homer do by default to avoid motif matches by change and since the actual TF binding (and therefore the motif) is a narrow event. Depending on your peak calling you might get peaks that are notably larger than 200bp and therefore scan way more nucleotides than necessary, increasing the change of false-positives.