Obstacle detecting method and system for automatic agricultural machine driving

Abstract

The invention discloses an obstacle detecting method for automatic agricultural machine driving. The method comprises: a binocular vision collector acquires image data containing distance information;the image data are inputted into a deep convolutional neural network, the deep convolutional neural network segments the image data into a plurality of sub images, each sub image is processed, objectblock information and object block type information that may be included by each sub image are predicted, and a credibility score of a predicted object is calculated based on the object block information and object block type information; and a credible identified object position coordinate and distance information are determined and outputted. In addition, the invention also discloses an obstacle detecting system for automatic agricultural machine driving. The detecting system is composed of an image data acquisition module, an identification module and an output module. According to the invention, the position information and distance information of the identified object are outputted based on the obtained image information, so that obstacles on the farmland are detected effectively; false background identification is reduced; and alarming is carried out based on a set dangerous area.

Description

technical field [0001] The invention relates to the technical field of information processing, in particular to an obstacle detection method and system for automatic driving of agricultural machinery. Background technique [0002] In the process of farmland operations, there are inevitably various obstacles in the field. If the obstacles cannot be detected, serious accidents will occur, causing casualties or economic losses. Therefore, it is necessary to effectively detect obstacles in the farmland . However, the identification of obstacles in the farmland has its specific difficulties and challenges. On the one hand, crops and obstacles often have similarities in color and texture, which poses a high challenge in the detection accuracy of obstacles; on the other hand, various objects in the field, including crops and fields Color and texture also change with the seasons; this makes obstacle detection even more difficult. [0003] An existing farmland obstacle detection m...

AI Technical Summary

Problems solved by technology

[0007] 1. Background misrecognition: The above detection method needs to rely on the difference in color and / or frequency between crops and obstacles for detection. When the similarity between the two is too high in color or frequency, it is not easy to detect obstacles , so the background becomes a great interference, causing background misrecognition

[0008] 2. Lack of detected object position and distance information: the above detection method only simply recognizes the object, does not give specific object position and distance information, and cannot effectively find obstacles and make accurate judgments

Images