Dirt self-checking method of digital camera image sensor based on retina perception

Abstract

The invention relates to a dirt self-checking method of a digital camera image sensor based on retina perception, which comprises the following steps of: 1, placing a pure color object in front of a digital camera lens, and photographing to obtain image data of a frame of RGB channel; 2, reading image data and carrying out graying by a priori method; 3, setting a convolution kernel, and carrying out convolution-based mean value downsampling; 4, performing convolution calculation and retina perception-based self-quotient image calculation to obtain data in a floating point form; 5, performing data processing; 6, performing normalized cross correlation template matching by adopting a normalized cross correlation algorithm to obtain a correlation heat map; 7, performing thresholding processing on the correlation heat map; and 8, analyzing the connected domain to obtain the dirty position of the digital camera image sensor. The method can quickly and effectively extract and detect the dirtcharacteristics of the digital camera image sensor, and is low in detection cost, high in processing speed and easy to implement.

Description

[0001] (1) Technical field [0002] The invention relates to a detection method of an image sensor, in particular to a method for self-detection of dirt of a digital camera image sensor based on retinal perception. [0003] (two), background technology [0004] Image enhancement methods are an essential image processing step in autonomous driving, outdoor robots, aerospace, biomedicine, public safety, AR, VR, and intelligent manufacturing in industry. Before the computer vision algorithm is formally processed, the image enhancement algorithm will be performed on the original image to improve some details in the image, such as tone remapping, edge enhancement, etc. This method is based on stretching the feature part with a small original value range into a larger dynamic range. [0005] Most real-time algorithms are based on artificial modeling, so the algorithm needs to perceive the global grayscale, such as grayscale thresholding, HOG, cluster analysis, FSAT features, etc. D...

AI Technical Summary

Problems solved by technology

Due to the need for image enhancement, when the camera sensor has been working for a long time in a poor working environment, when the sensor adheres to small debris that has fallen into the body, the dark spots of the image will be further highlighted with the enhancement process, which directly affects the algorithm. reliability and even availability

[0006] Dirt on camera sensors is a hidden danger of all high-precision, high-tech vision-based smart equipment. For obvious dirt, human eyes or traditional detection algorithms can be used to identify whether the sensor is affected by dirt, and some light and small Dirt can affect the visual system more. Usually, the more obvious dirt will be processed to get very deep dark spots, which are easy to identify, while the lighter dirt will be mixed in the grayscale features of the collected objects, causing detection problems. , tracking and other algorithms failure or instability

[0007] There are generally two types of dirt detection technologies for existing camera sensors. One is to take pictures of the surface of the sensor to be detected by peripherals. Because the general detection environment cannot meet dust-free conditions, this method is a theoretical method and has little practical application significance; The other is the self-inspection technology that does not need to be disassembled. The recognition rate of this technology is low, mainly because the imaging principle of the lens will cause uneven exposure and vignetting phenomenon, showing the effect of bright center and dark surrounding, which cannot pass the threshold. The optimization operation can effectively segment the dark spots that are not obvious

[0008] Human vision has complex self-adaptive characteristics, which are mainly reflected in the fitting or filtering characteristics of the slowly changing lighting environment. Therefore, when using the manual detection scheme, most of the dirt can be observed through some enhanced algorithms. However, it is difficult to perform effective automatic detection directly on the enhanced result, mainly because different types of light sources cannot overcome the vignetting of the image in the incident lens, and the enhanced result still retains the vignetting effect (such as figure 1 shown)

Images