The conventional single-wavelength laser wireless power transmission (LWPT) system has some problems, such as low photovoltaic conversion efficiency, excessive concentration of light spots, and complex tracking-aiming system. In this research, a novel LWPT system based on deep learning algorithm is proposed. The white laser is used as the energy transmission source and the security camera is used as the visual recognition source. A new fluorescent ceramic heat sink structure is proposed. This let 60 W white laser lamps work steadily at 70 ℃ ambient temperature. The light intensity is more than 6 000 000 cd, the luminous half angle is 1.8°, and the light spot covers the target more evenly. An algorithm based on YOLOv5 for recognition and target posture acquisition is presented. After training, the mean average precision (Intersection over Union 0.5) of the model is 99.78%. Within the 8m range, the 60 W white laser lamps illuminate solar cells to power up the LED lamps. Outdoor test shows that the novel LWPT system can automatically recognize, track, and illuminate solar cells which are carried by the UAV (unmanned aerial vehicle). The system is more robust to avoid street lamp interference.