Research On Laser Active Detection And Recognition Technology Of Concealed Photoelectric Target

With the popularization of photoelectric device such as burglary pinhole cameras and miniature cameras, indoor activities are facing an increasingly serious threat of information disclosure. The leaked sensitive information may bring serious military, social and business losses. In order to solve this problem, laser active detection and recognition technology can be utilized detect interior hidden reconnaissance equipment as an anti- reconnaissance method, and eliminate potential threats.The laser active detection and recognition technology mainly uses the strength characteristics of target’s echo signals, based on its “cat’s eye effect”, for identification. However, in the indoor environment, due to the influencing factors of short operation distance and high reflectivity objects, some disturbing objects shall reflect strong signals corresponding to the interrogating laser beam. The reflected signals, in some cases, are very close to the true target echo signals on the strength characteristics. Therefore, it is unable to identify the true targets only depending on the strength characteristics of the "cat’s eye effect". This thesis deeply studied the features and characteristics of photoelectric systems under the laser interrogation for detection and recognition, solved the problems of technology application in indoor environment and developed a hand-held hidden camera reconnaissance equipment.This study presents a local texture criterion for describing the distribution characteristics of the laser cross section in order to overcome the first problem. A comprehensive criterion combining the local texture criterion and a modified version of the morphological criterion has also been presented for identifying the true target against false positives. The average recognition rate of the comprehensive criterion is 91%, which are increased by 23% and 23% respectively comparing with single use of texture or morphological criterion.To counter the problem of real-time detection and identification process, this article mainly solve it from the perspective of algorithm optimization. We present a connected domain search method based on concentric dual window template matching algorithm to obtain positions of the reflection signals in the image. The algorithm introduces morphological constraints in the process of searching and rules out regions significantly different from those of true targets on the morphological characteristics. By this, it reduces calculation of unnecessary region labelling and improves running speed. The embedded system designed for the prototype is using DSP + FPGA architecture and has been optimized for boosting running speed. The experimental results show that the average PC computing time is 0.0804 s and that of the prototype is 0.097 s, as to meet the requirement of real-time process. Compared with the recent methods, the algorithm on calculating speed is 19.1-23.0 times of the MSPF algorithm in parallel mode; 12.4-15.0 times of the MSPF algorithm in serial mode; 35.4-42.7 times of the SPCS algorithm; and 63.7-76.9 times of the SFDC algorithm.As for the technology implementation on the development of practical portable system, this paper puts forward an adaptive threshold segmentation algorithm based on the sparsity of targets. It, cooperating with narrowband filter design, solves the coarse segmentation between the photoelectric target retro-reflection signals and background reflection signals within the field of view. The method replaces the conventional active and passive image difference strategy and overcomes the defects of implementation on hand-held operation type equipment.A large number of experiments have been performed to analyze and evaluate the performance of all aspects of the proposed recognition method and system. The PC experimental results indicates feasibility, adaptability, real-time and limitations. Feasibility study demonstrated the correctness on the principles of the proposed criterion and algorithm. Adaptability research shows that the algorithm is robust to the outside light intensity and is not sensitive to the operating range within the scope of 2.2-4.2m. The experimental results of real-time have been mentioned before. Research of limitations shows that when the scene contains high reflectivity objects such as polishing spherical distractors or interfere with circular metal objects, and when their reflected signals have similar intensity distribution properties with the true targets, the proposed method this paper fails.This thesis made a pioneering work on applying laser active detection and recognition technologies of photoelectric target in the indoor environment, made contributions to the characteristics and features for identifying photoelectric target accompanied with the measurement criteria, and developed prototypes with the ability to practical application. The work filled the blank of the domestic research in this field and practical. It also activated the active detection and recognition technologies for photoelectric equipment reconnaissance.