Sometimes when we get an optical system design request, we may see the following criteria: target detection distance: ≥12 km and target recognition distance: ≥5 km. How does this index correspond to the parameters of the optical system? This paper will introduce this problem.
General principle of target search
Let’s assume that the human eye is looking at an image. When the human eye is searching for a target in a fixed background or an image on an imaging system display, the continuous response of the eye is divided into four levels of detection (discovery) , classification, recognition and discrimination, or the process of searching for objects by the human eye consists of this interrelated step. The meaning of these steps can be summarized as follows:
1.Detection:
Presence of a target: the primary task of detection is to determine the presence of a target in the field of view. This is usually done by detecting contrast or other salient features between the target and the background.
Signal Processing: the use of filtering, enhancement and other techniques to improve the visibility of the target, so that it stands out from the background.
2. Classification:
Object classification: according to the shape, size, motion characteristics of the target, will be classified as a certain class of objects (such as people, vehicles, animals, etc.) .
Feature extraction: The key features of the target are extracted to help the classification algorithm to identify the target class more accurately.
3.Recognition:
Detailed classification: The Identification Stage further classifies the target into specific subcategories (e. g. trucks, cars in cars, etc.) .
Pattern recognition: using template matching, machine learning and other technologies, according to the extracted features of pattern recognition, so as to determine the specific type of target.
4. Identification:
Detail recognition: on the basis of recognition, further distinguish the specific model, type or individual characteristics of the target.
High Resolution Imaging: the imaging system is required to have a higher resolution in order to capture the details of the target features.
In the process of target detection, classification, recognition and discrimination, people experience subjective judgment, so there is a problem of search probability in target search. The search probability provides an evaluation of the performance of the photoelectric imaging system to a certain extent. Of course, the search process is very complex, the search probability is related to certain tasks (detection or recognition) , but also with the observer’s state, background characteristics and so on.
The Johnson criteria
The different stages of target search process correspond to different detection level, which is a kind of evaluation method that combines the system performance with human vision.
During the 1950s, cold war-era demand for infrared imaging systems and night vision equipment increased in the United States. At the time, researchers needed a quantitative method to assess and compare the performance of different imaging systems to ensure that they met military and security needs.
The Johnson principle was developed by American scientist Johnson in his work on infrared imaging systems. His work focuses on how to effectively evaluate the performance of imaging systems for different tasks, such as detection, identification, and discrimination. Johnson linked the target detection problem to the equivalent strip pattern detection problem. He found that in engineering practice, it is possible to determine the target recognition ability of an imaging system by the resolution of the target equivalent strip pattern without considering the nature of the target and image defects.
The equivalent strip pattern of a target is a set of strip patterns with equal spacing between black and white. The total height of the strip pattern is the critical size of the target that can be basically recognized, that is, the minimum projection size of the target, the strip length is the size across the target perpendicular to the critical dimension direction. The discernibility of the equivalent strip pattern is the number of discernible strips contained in the target critical size, usually expressed as“Perimeter/critical size”.
The Johnson criteria provide a quantitative assessment for each of these stages, specifying the spatial resolution required by the imaging system to accomplish these tasks. The core idea is as follows:
1. Period target width: Johnson’s criterion measures the resolution by the spatial frequency (period) of the light and dark changes in the target image. A“Cycle” represents a complete light-dark change in the target image.
2. Resolution Requirements:
Because of the above detection level of the required number of band periods are obtained in 50% probability. Therefore, in using Johnson’s criterion, we may need to modify the number of band periods corresponding to other probabilities.
