For autonomous robots, small target motion detectionin complicated natural environments is an extremely challenging task. The extremely efficient visual systems of insects detect mates and trackprey, though the target occupy minute degrees of its visual field. Small target motion relies on the excellent sensitivityof specialized neurons, called Small TargetsMotion Detector (STMD). The already existing models based on STMD, hugely depend on visual contrast butshows poor performance in complicated natural environments, whereas small targets commonly exhibit highly low contrast against its backgrounds nearby. Here, we frame an attentionpredictionguided visual system to break the limitation. It hasfive main subsystems, a) Acquisition Moduleb) Process Modulec) Attention Module d) Action Module e) Log Module. These Modules Capture the image when an abnormal motion occurs and then enhanced the image, stores. The image gets validated and sends to the relevant destination receiver.detailed experiments on various real-world datasets prove the efficiency and majority of this system in the detection ofminute, lowcontrast movements of targets against complicated natural environment. This method is more efficient andproductive because the recording and storage take place when the abrupt changes occur. It is more cost effective.