Method and device for measuring size of target

Abstract

One technical scheme of the invention is to provide a method for measuring the size of a target. Another technical scheme of the invention is to provide a device for measuring the size of a target. According to the invention, deep learning and computer vision technologies are utilized to dynamically identify the states of the reference object and the measurement target, and the states of the reference object and the measurement target are automatically measured and recorded. For example, the sizes of reference objects (such as biopsy forceps), background targets (such as normal target tissuesin backgrounds of colon bands, trilobal gates, pentenos, duodenal descending papilla, ileocecal valve and like) and / or abnormal targets (such as abnormal target tissues of flat polyp, abnormal blood vessel fractal, atypical hyperplasia, inflammation, bleeding, excrement residues, tumors, progressive cancer and like) in endoscopic examination process are judged, and accurate and specific numericalvalues are outputted.

Description

technical field [0001] The invention relates to a method for measuring the size of an object and a device for measuring the size of an object, especially suitable for measuring the size of an object under an endoscope. Background technique [0002] Existing abnormal object identification methods mainly include: 1) Human identification: the size judgment depends on the physical strength, attention, memory, inspection time and operating experience of the operator. The size judgment of the report is not accurate, and the judgment of most operators is not accurate enough. The sizes of anomalous targets reported are estimates. 2) Physical identification: through laser light source or laser radar, the time delay of pulse emission and reflected echo is proportional to the imaging distance, and the time delay of reflected echo pulse is measured to obtain the imaging distance, and then obtain the size of the lesion. However, such methods Additional pulse emission and detection units...

AI Technical Summary

Problems solved by technology

[0002] Existing abnormal object identification methods mainly include: 1) Human identification: the size judgment depends on the physical strength, attention, memory, inspection time and operating experience of the operator. The size judgment of the report is not accurate, and the judgment of most operators is not accurate enough. The sizes of anomalous targets in the report are estimates

2) Physical identification: through laser light source or laser radar, the time delay of pulse emission and reflected echo is proportional to the imaging distance, and the time delay of reflected echo pulse is measured to obtain the imaging distance, and then obtain the size of the lesion. However, such methods Additional pulse emission and detection units are required, the system structure is complex and the cost is high

3) Software recognition: It can only calculate the size of the reference object of the overlapping bounding box, but cannot judge the distance between the two, and the size judgment accuracy is not enough. Even if the deep learning technology is used in the current monocular camera, it relies entirely on computer vision technology. In a non-geometric environment Among them, identifying the size of non-Lambertian bodies is also a challenging task; fully automatic judgment, there is no human experience and judgment in the process, and the combination of man and machine has not been achieved, which is not rigorous enough as a medical device; the goal can only be Specific lesions, which cannot be selected during the examination

Images