Identification and Interpretation of Martian Craters and Climate History, 15-9311Printer Friendly Version
Inclusive Dates: 05/07/02 - 09/05/03
Background - Mars exploration is slated to be the highest priority element of NASA's Solar System Exploration Program for the next two decades. Much of the basis for this interest in Mars relates to geological evidence for running water on, or just below, the surface of that planet, especially several billion years ago when the Red Planet's climate might have been warmer. The implications for past or present life are obvious. The abundant impact craters on the southern highlands of the planet, among which the ancient river valleys weave, provide witness to the evolving surface processes on Mars as their initial fresh bowl shapes become degraded by erosion and in-filling. We proposed to develop scientific and technical skills as well as image processing capabilities in Divisions 10 and 15 in order to prepare us to research these vital Mars science issues using the vast image data sets collected by past and ongoing spacecraft, especially Mars Global Surveyor (MGS) and Mars Odyssey. In our first and largest task, the P.I. and all Co-I's, assisted by Mr. Enke in the Boulder Office, extended our previous automated crater-detection capability (developed by Div. 15 in collaboration with J.P.L.) to a more complex Martian terrain, with craters and superimposed drainage channels, by developing several enhancements. In addition, Div. 10 expertise was utilized to modify and apply several different machine vision techniques to the same study area on Mars. In our second, smaller task, P.I. Chapman investigated several scientific issues associated with Martian cratering, focussing on investigating the morphological attributes of Martian craters to test an important hypothesis concerning a warm, wet period in the history of the Red Planet immediately following the Late Heavy Bombardment about 4 Gyr ago.
Approach - In addition to testing and improving the algorithms previously developed with J.P.L. colleagues, and applying them to more complex images of Mars, we evaluated several independent and/or different algorithms for identifying Martian craters and rejecting non-crater features. These were: (a) cross-correlation based template matching, (b) directional edge based detection, (c) convolution with an annular crater kernel, and (d) circular Hough Transform. Each approach has its advantages and disadvantages in detecting craters of certain sizes and morphologies. The overall approach to the project involved not only cross-disciplinary studies but also long-distance interactions between investigators in San Antonio and Boulder, facilitated by rapid exchange of visual materials on the Internet. The smaller task involved a collaboration with a busy non-SwRI researcher, which largely did not come to pass. Hence efforts were partly diverted to studying interesting new issues involving secondary crater formation on Mars.
Accomplishments - We discovered that the most successful approach to automatic recognition of Martian craters was to apply the variety of different algorithms (each with rather strict, high-confidence parameters) and then to combine the results. This not only maximizes detection of impact craters on planetary images but also minimizes false detections. We expect that our skill in using these automated approaches has already reached a level that would permit our approach to be used as an "assistant" to a human crater measurer to vastly increase that person's efficiency. The analyst would only have to correct and augment the identifications and measurements rather than to identify all the craters from scratch. We also made some insights concerning Martian crater morphologies and the role of secondary cratering on Mars.