General object real-time learning and recognizing system and learning and recognizing method thereof

Abstract

The invention discloses a general object real-time learning and recognizing system and a learning and recognizing method thereof. The system comprises an error correction module, an organization encoding module, a deep learning module, a database management module and an image processing module. The method comprises the following steps: firstly, acquiring and processing image data, detecting position information of an object in an image, acquiring a label box, and then extracting, encoding, organizing and retrieving color blocks and color block characteristics of the image in the box so as toidentify the object. The obtained image is segmented into simple shape combinations based on specific image features and a binary encoding mode, and the extracted features are classified and refined,so the generalization and promotion capabilities of the system are improved; after a new object image classification is obtained each time, the features can be continuously extracted for learning, recognition is performed in the subsequent process, and a large number of samples do not need to be manually obtained for long-time training, so finally continuous tracking, learning and recognition canbe performed on the object in the image in real time.

Description

Technical field [0001] The invention relates to the technical field of image recognition, in particular to a universal object real-time learning and recognition system and a learning and recognition method thereof. Background technique [0002] Image recognition technology refers to the object recognition of images to identify targets and objects in various patterns. The development of image recognition has gone through three stages: text recognition, digital image processing and recognition, and object recognition. The research on text recognition began in 1950, generally for the recognition of letters, numbers and symbols, from printed text recognition to handwritten text recognition, the application is very wide; the research on digital image processing and recognition is very early, and it has a history of nearly 50 years. Compared with analog images, it has great advantages such as storage, transmission, and compression, less distortion during transmission, and convenient p...

AI Technical Summary

Problems solved by technology

[0006] 1. A large number of sample images are required for long-term model training, which is time-consuming and troublesome;

[0007] 2. It can only recognize the trained object classification, which is highly targeted and poor in scalability, that is, it cannot recognize any object, and the recognition range is only within the sample range. If you need to recognize a new object type, you need to re-do the object type sample It takes a lot of training to achieve;

[0008] 3. The recognition accuracy depends on the sample. If the sample does not contain a certain object feature, the accuracy of recognizing the object will be reduced;

[0009] 4. Too many types of object samples will lead to a decrease in recognition accuracy;

[0010] 5. The algorithm process cannot be explained, for example, when you put an image of a cat into a neural network and the prediction result shows that it is a car, this result cannot be explained

Images