Recognition of low-light level target is attracting more and more concern in modern military areas. However, for the
reason of low contrast, low signal-to-noise ratio and inadequacy information of low-light level target etc, the goal to
detect and recognize the target would not be realized by using photoelectric joint transform correlator. By median
filtering and edge detection with lifting wavelet transform for low-light level target in this paper, the interference of
background noise is reduced and useful information of target and template is enhanced at the same time. Experimental
results show that the brightness and contrast of correlation peaks are both improved obviously after processing the joint
image, which proves the method is very effective in target recognition field by using photoelectric hybrid joint transform
In this paper, we used hybrid optoelectronic joint transform correlator (HOJTC) for recognition of moving target.
HOJTC is one of the most successful optical pattern recognizer which is composed of laser, spatial filter, Fourier
transform lenses, EALCD (electrically addressed liquid crystal displays) and CCD (Charge Coupled Device). It has
many applications in the fields of industry and military affairs.
The speed of armored vehicle is generally less than 80 km/h. CCD used as receiver can capture 25 frames per second.
The difference caused by shape, scale and rotation always exists between template and target. Therefore, the optical
correlator can only detect captured moving target about 3 to 5 serial frames. For some targets in cluttered scene, it even
can not recognize the target, which means the tracked target is missing. It shows the influence of clutter and distortion
brings great difficulty to correlation recognition. In order to realize scale invariable and rotation invariable, the method of
adaptive threshold is applied. After processing the images of moving targets, we can reduce the influence of cluttered
background. The effect brought by the changes of shape, scale and rotation is also reduced. Consequently, the ability of
automatic recognition, location and tracking of moving target by HOJTC can be enhanced greatly.
The experiments are performed to recognize moving tanks with high speed about 70 km/h. The experiments show that
more than 80 serial frames can be recognized after target images are processed. The joint transform correlator can
recognize even more than 150 frames when the target is in relative clean scene. It has great meaning for target detection
and tracking. The conclusion can be drawn that the proposed method of adaptive threshold for moving target is feasible,
and it could effectively enhance the ability of automatic recognition and tracking of a moving target.
Hybrid optoelectronic joint transform correlator (HOJTC), exploiting the Fourier transform property of a lens, implements target detection in real time. Adaptive nonlinear digital filtering in the joint transform power spectrum (JTPS) plane improves the immunity against the noise and clutter. In this paper, electrically addressed liquid crystal devices (EALCD) are used as the space light modulators (SLM), Charge Coupled Device (CCD) matrix camera as the square law detector and Ar+ laser as the light source. We develop the hybrid optoelectronic joint transform correlator controlled by computers, which can successfully detect and recognize the target in cluttered scenes in real time. The speed rate of the recognition is 25 frames per second. As the experiment examples, the target recognition of a tank and a car in cluttered scenes is presented. The experiments show that the intensity of cross-correlation peaks after adaptive nonlinear digital filtering is increased greatly, the performance of the joint transform correlator is improved, and it can detect the distorted and noisy targets in cluttered scenes in real time. It was found that using wavelet filter in frequency domain is a very effective way to suppress clutter noise while maintaining high tolerance for distortion.