Intensive target detection metering method

Abstract

The invention relates to the technical field of image recognition, and aims to provide an intensive target detection metering method. The method comprises the following steps: inputting a to-be-detected original image into an intensive target detection model; enabling the intensive target detection model to position a target area in the original image, and then output a bounding box of the targetarea; cutting the original image according to the bounding box of the target area to obtain a target image and positioning information of the target image, and inputting the target image into a classification model; enabling the classification model to perform image classification on the target image to obtain category information of the target image; and integrating the positioning information and the category information of the target image, and filtering redundant images in the target image to obtain the positioning information and the category information of the intensive target. Accordingto the method, required training samples are reduced, the acquisition cost is reduced, and meanwhile, rapid iterative updating can be realized.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to an intensive target detection and measurement method. Background technique [0002] Intensive target detection tasks, compared with general target detection tasks, are difficult in that the number of objects in the image is very large, ranging from dozens to hundreds. Objects are close together, requiring specific adjustments to general object detection methods. Taking panoramic shelf recognition as an example, panoramic shelf recognition is to shoot multi-layer (4 floors and above, 2.5m and below) shelves, and use deep learning methods to identify product categories and positioning on the shelves. In the panoramic shelf scene, nearly 100+ products in the front row are closely distributed together, and at the same time, they are also closely connected with the same products and the same color. In this scenario, the image captured by the camera will have a product with...

AI Technical Summary

Problems solved by technology

However, in the existing technology, the general target detection training method is usually used. When the detection target is updated, a large number of training samples are required to retrain the detection model, which is expensive to collect and is not conducive to rapid iterative update.

Images