Visual servo laser positioning system and method of picking robot

Abstract

The invention provides a visual servo laser positioning system of a picking robot and a corresponding method. A controller trains an existing depth learning network algorithm, a monocular camera acquires a two-dimensional image of a picking environment, inputs the two-dimensional image into the algorithm to identify a target object and outputs direction angle information relative to the monocularcamera, a two-dimensional holder points to the direction of the object to be detected, and a laser ranging module is used for ranging; and finally, the controller calculates the actual spatial information of the object to be detected according to the measurement result. According to the system and the method, the advantages of depth vision and laser ranging are combined, monocular depth vision andthe single-point laser ranging module with a two-dimensional servo structure are adopted to measure the spatial positions of a limited number of targets in an outdoor environment, so that the required specific target objects are spatially positioned in an unstructured environment, the measurement precision can be effectively improved, the data processing amount is reduced, the cost is reduced, and the system and the method are suitable for picking various fruits and vegetables in indoor and outdoor environments.

Description

technical field [0001] The invention belongs to the field of object recognition and positioning, in particular to a visual servo laser positioning system and method for a picking robot. Background technique [0002] With the gradual improvement of the modernization of agricultural production, more and more traditional human labor can be replaced by machines. Picking is the most time-consuming, labor-intensive, and seasonal link in the agricultural production chain. Therefore, improving the automation of picking will greatly improve the efficiency of agricultural production. Automatic picking will inevitably involve the identification and positioning of harvestable parts such as fruits on crops. The existing methods for spatial positioning of objects include: binocular vision, laser radar, structured light camera and TOF camera. Among them, based on the binocular vision of the detection point matching principle, the target is positioned by the method of triangulation measure...

AI Technical Summary

Problems solved by technology

Among them, based on the binocular vision of the detection point matching principle, the target is positioned by the method of triangulation measurement. In this method, the corner point data is too sparse or too dense, and the requirements for the corner point matching algorithm are very high. Accurate acquisition The spatial position of the object is more difficult; the surface scanning laser radar is not suitable for the spatial positioning requirements of small objects due to its low scanning resolution; both the structured light camera and the TOF camera measure by actively emitting auxiliary light sources, and the structured light The coded infrared light emitted by the camera is positioned by identifying the code change of the infrared light projected on the surface of the object, while the TOF camera emits infrared light with certain information, which is determined by calculating the time it takes for the light to reflect on the surface of the object. Realize the acquisition of the distance of the object, but these two methods are easily interfered by the sun's rays in the outdoor environment due to the active luminescence, so that the accurate information of the object cannot be obtained

Images