Abstract: A method for orchard autonomous basket searching based on VINS-MONO and improved YOLO v4-Tiny was proposed for the scenario of orchard transporter fruit harvesting and autonomous transportation operations. Firstly, the position of the orchard transporter was estimated in real time based on the VINS-MONO visual inertial odometry calculation method. Then based on the improved YOLO v4-Tiny target detection algorithm, real-time target detection of the fruit basket was performed on the image data, and its depth information was obtained. Nextly, the position of the identified fruit basket was estimated based on the current position pose of the transporter, the depth information of the fruit basket and the parameters of the depth camera. Finally, based on the principle of cubic B spline curve fitting, the path of the identified fruit basket was fitted in real time, using the position of the identified fruit basket as the control point, to provide route guidance for the autonomous fruit basket search of the orchard transporter. The test results showed that the average recognition accuracy of the improved YOLO v4-Tiny fruit basket recognition model was 93.96%, the average inference time was 14.7 ms, the measurement error of the fruit basket distance within 4 m was less than 4.02%, the measurement error of the fruit basket angle was less than 3°, the measured driving speed of the orchard transporter was 3.3 km/h, the average update time of the fruit basket search route was 0.092 s, and the transporter could stably achieve autonomous basket searching in two operating environments: fruit tree rows and orchard roads. This method could provide autonomous basket finding path guidance for orchard transporter and provide research reference for visual navigation of orchard transporter.