Extraction Of Mineral Alteration Information Based On Ant Colony Optimization Algorithm And Support Vector Machine

Mineral alteration zones are delineated by ground survey in traditional methods of geological ore-searching. Mineral alteration data are extracted and spatial distribution features of altered rocks are acquired by remote sensing information extraction ways. Although remote sensing method is superior to ground survey in traditional methods in saving manpower and resources, common remote sensing information extraction methods can't extract the mineral alteration data well because the mineral alteration data are too weak to be extracted in remote sensing data. It has great practical significance to study on finding new remote sensing altered information extraction methods in order to enhance the reliability and economic benefit of remote sensing exploration.Ant Colony Algorithm (ACO) is a novel biomimetic algorithm which stimulates collective routing in nature and has some good features, such as discreteness, parallelism, robustness, positive feedback, etc., and is recognized and accepted rapidly for its advantages, such as explicit concepts, convenient realization. Support vector machine (SVM) is a kind of novel machine learning methods, based on statistical learning theory, which has become the hotspot of machine learning because of its excellent learning performance, and it has been applied in many domains successfully.This paper is supported by the project: Study on the technology of remote sensing comprehensive information extraction of mineral resources and the application to ore-searching, which is the subject of the project of Great Geological Survey of Land and Resources: Study on the Technology of Whole-Band Quantitative Remote Sensing and Its Application in Resource and Environment Survey (Contract serial number: 1212010660601).Tongren-Zeku area, belong to Huangnanzhou city of Qinghai Province, was chosen as the study area, and the study of mineral alteration information was extracted based on Ant Colony Algorithm and Support vector machine.The main conclusions are listed as follows:(1) The adopting parameters in the classifier model of SVM can make significant influence to the property of the classifier. ACO algorithm was selected to be used in the main parameters searching in order to avoid time consuming and scoping area uncertainty duo to grid searching method. ACO algorithm is more quickly and better than grid searching in finding out main parameters of SVM, through simulation test.(2) SVM mineral information extraction model, based on principal component analysis, was established. In order to solve the defaults of traditional mineral alteration information extraction, such as large sample volume and difficulties of sample selection, a new extraction model, based on principal component analysis and SVM, is proposed in this paper. Using this new model, not only can solve the defaults of traditional extraction method, but also can ensure the extraction precise of mineral information. Through the validation on the spot and overlay analysis with known samples, it has been validated that this is an effective way to extract alteration information.(3) The SVM mineral information extraction method, based on texture and spectra, is proposed in this paper, which both overcome the limitation of information content by combining pixel spectrum and spatial texture and structure, as well as combining spectrum resolution and spatial resolution of remote sensing data, and ensure the mineral sample precise of SVM. Overlaying the extracted remote sensing anomaly alteration information with the known mineral field, the effect is good. Through the validation on the spot, alteration phenomenon in different degrees has been found, and three key anomaly areas were pointed out.(4) The spectral unmixing method based on ACO algorithm, which is the first time to import ACO algorithm to remote sensing geology field, is proposed to reject vegetation disturbance information. This method makes the target recognition quickly, and has been proved to be doable to reject vegetation information by error diagram analysis and contrast analysis between original map and the rejected vegetation image.(5) The argillization alteration extraction information model, combining ACO algorithm and ratio method in traditional remote sensing geology area is firstly established. Awa area, belong to Huangnan city of Qinghai province, was selected as study area. First, the rules of hydroxyl alteration information extraction model were established. Then, hydroxyl alteration information extraction model, based on ACO algorithm, was established. Finally, hydroxyl alteration information was extracted by the model. Overlaying the extracted remote sensing anomaly alteration information with the known mineral field, the effect is good. Through the validation on the spot, alteration phenomenon in different degrees has been found.