Commodity invalidation detection and judgment scheme based on position of container door

Abstract

The invention discloses a commodity invalidation detection and judgment scheme based on the position of a container door, and the scheme comprises the steps: capturing an image: carrying out the imagecapturing of a container door region through a fisheye lens, and obtaining a detection image of the container door region; image extraction: intercepting an effective area of the captured image, andremoving an invalid area in the image; wherein the square container door is distributed in a fan shape above the images, and the rough position of the fan-shaped area of the container door is obtainedaccording to the multiple images; edge correction: performing accurate correction of image edges according to pixel histogram features and image filtering; and position determination: accurately positioning the accurately corrected image to obtain the accurate position of the glass door. The method is reasonable in design, can accurately detect the area of the container door, and can also avoid accumulated errors generated in the subsequent detection process at the same time.

Description

technical field [0001] The invention relates to the technical field of commodity detection, in particular to a scheme for detecting and judging commodity invalidity based on the position of a container door. Background technique [0002] The existing unmanned container algorithm will detect the whole image indiscriminately in the detection process. In addition, the reflection of the container door in the fisheye lens or the transmission of objects outside the door will interfere with the original product detection and recognition model, resulting in false detections and ultimately affecting the accuracy of product detection and recognition. [0003] In order to solve this problem, the existing technology collects and marks the sub-pictures belonging to the container door in the picture as negative samples, and inputs them into the detection and classification network together with the marked positive samples for training. This approach, in addition to bringing additional la...

AI Technical Summary

Problems solved by technology

More importantly, the introduction of too many negative samples will have a negative impact on the detection model, for example, it will cause positioning or category errors for the products that were originally detected correctly. This error will gradually accumulate in the subsequent modules of the container system, and will eventually The detection and identification results have adverse effects

Images